Laminar is an observability platform for AI agents. Every tool acts on one Laminar project. With a project API key the project is fixed. When signed in with OAuth, call `list_projects` and pass the project's id as `projectId` to every other tool.
Discoverability
0/6 passed0 of 6 passed0/100
Client listings and public setup information.
mcp.json or server-card.json
Either mcp.json or an MCP Server Card declaring this endpoint passes.
How to fix: For Claude directory submission, declare the applicable annotations.readOnlyHint or annotations.destructiveHint value as a boolean on every tool.
How to fix: On each tool, declare annotations.readOnlyHint, annotations.destructiveHint, and annotations.openWorldHint as booleans. The policy review separately assesses whether their values match the tool's actual behavior.
How to fix: Set the top-level tool.title to a short human-readable label (for example, "Search catalog"). Claude directory submission requires tool titles; annotations.title remains accepted only as a legacy fallback.
The title clearly communicates the tool’s action and subject. Keep it concise.
Description74/100
Add tags to a trace, e.g. to mark traces you reviewed. Tags are queryable with
`query_laminar_sql` (`traces.tags`).
The description explains the capability, when to use it, and what to expect.
Inputs58/100
2 of 3 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
traceId
Not provided
Missing description. Explain what this value means and how to supply it.
tags
Tag names to add; tags the trace already has are kept.
Clear enough to supply this argument.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "add_trace_tags",
"description": "Add tags to a trace, e.g. to mark traces you reviewed. Tags are queryable with\n`query_laminar_sql` (`traces.tags`).",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"traceId",
"tags"
],
"properties": {
"tags": {
"type": "array",
"items": {
"type": "string"
},
"description": "Tag names to add; tags the trace already has are kept."
},
"traceId": {
"type": "string",
"pattern": "^(?:[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}|[0-9a-fA-F]{32})$"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description90/100
Ask Laminar Agent a question about this project's traces in plain language and get a
grounded answer. The agent runs read-only SQL and reads traces itself. The answer ends
with a `conversationId`; pass it back to continue the conversation.
The description explains the capability, when to use it, and what to expect.
Inputs67/100
2 of 3 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
prompt
Not provided
Missing description. Explain what this value means and how to supply it.
conversationId
Server-minted id to continue a prior conversation. Omit to start fresh; the response returns
the new id to pass back on the next call.
Clear enough to supply this argument.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "ask_agent",
"description": "Ask Laminar Agent a question about this project's traces in plain language and get a\ngrounded answer. The agent runs read-only SQL and reads traces itself. The answer ends\nwith a `conversationId`; pass it back to continue the conversation.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"prompt"
],
"properties": {
"prompt": {
"type": "string"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
},
"conversationId": {
"type": [
"string",
"null"
],
"default": null,
"description": "Server-minted id to continue a prior conversation. Omit to start fresh; the response returns\nthe new id to pass back on the next call."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description70/100
Create a workspace LLM profile. Secrets you pass are stored encrypted and never returned.
The description explains the capability, when to use it, and what to expect.
Inputs9/100
2 of 17 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
name
Not provided
Missing description. Explain what this value means and how to supply it.
provider
Wire names shared with the frontend (`LLM_PROFILE_PROVIDERS`).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
$defs › ProfileAuth › oneOf › 0 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileAuth › oneOf › 1 › accessKeyId
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileAuth › oneOf › 1 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileAuth › oneOf › 2 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfig › region
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfig › resourceId
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfig › baseUrl
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfig › apiVersion
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfig › headerNames
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileSecrets › apiKey
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileSecrets › secretAccessKey
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileSecrets › token
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileSecrets › headers
Not provided
Missing description. Explain what this value means and how to supply it.
models
Not provided
Missing description. Explain what this value means and how to supply it.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "create_llm_profile",
"description": "Create a workspace LLM profile. Secrets you pass are stored encrypted and never returned.",
"inputSchema": {
"type": "object",
"$defs": {
"ProfileAuth": {
"oneOf": [
{
"type": "object",
"required": [
"type"
],
"properties": {
"type": {
"type": "string",
"const": "api_key"
}
}
},
{
"type": "object",
"required": [
"type",
"accessKeyId"
],
"properties": {
"type": {
"type": "string",
"const": "aws_keys"
},
"accessKeyId": {
"type": "string"
}
}
},
{
"type": "object",
"required": [
"type"
],
"properties": {
"type": {
"type": "string",
"const": "bearer_token"
}
}
}
]
},
"ProfileConfig": {
"type": "object",
"properties": {
"auth": {
"$ref": "#/$defs/ProfileAuth",
"default": {
"type": "api_key"
}
},
"region": {
"type": [
"string",
"null"
]
},
"baseUrl": {
"type": [
"string",
"null"
]
},
"apiVersion": {
"type": [
"string",
"null"
]
},
"resourceId": {
"type": [
"string",
"null"
]
},
"headerNames": {
"type": "array",
"items": {
"type": "string"
}
}
},
"description": "Non-secret, provider-specific part of `llm_profiles.config`. Fields are\npermissive on purpose: `service` validates on write, and a stale or\nhand-edited row should fail with a readable `ConfigError`, not a serde error.\nAbsent options are omitted on the wire because the Zod schema rejects `null`."
},
"ProfileSecrets": {
"type": "object",
"properties": {
"token": {
"type": [
"string",
"null"
]
},
"apiKey": {
"type": [
"string",
"null"
]
},
"headers": {
"type": "object",
"additionalProperties": {
"type": "string"
}
},
"secretAccessKey": {
"type": [
"string",
"null"
]
}
},
"description": "Decrypted contents of `llm_profiles.secrets`; also the plaintext shape the\nwrite routes accept. Absent keys are omitted on the wire (Zod rejects `null`)."
},
"LlmProfileProvider": {
"oneOf": [
{
"enum": [
"openai_completions",
"openai_responses",
"anthropic",
"gemini",
"groq",
"mistral",
"bedrock",
"azure_chat_completions",
"azure_responses",
"azure_anthropic"
],
"type": "string"
},
{
"type": "string",
"const": "custom",
"description": "OpenAI-compatible gateway over Chat Completions."
},
{
"type": "string",
"const": "custom_responses",
"description": "OpenAI-compatible gateway over the Responses API."
}
],
"description": "Wire names shared with the frontend (`LLM_PROFILE_PROVIDERS`)."
}
},
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"name",
"provider",
"models"
],
"properties": {
"name": {
"type": "string"
},
"config": {
"$ref": "#/$defs/ProfileConfig",
"default": {
"auth": {
"type": "api_key"
}
}
},
"models": {
"type": "array",
"items": {
"type": "string"
}
},
"secrets": {
"$ref": "#/$defs/ProfileSecrets",
"default": {}
},
"provider": {
"$ref": "#/$defs/LlmProfileProvider"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description82/100
Create a signal: an LLM judge that runs `prompt` on matching traces and extracts a
payload matching `structuredOutput`. Three things decide when it runs: `trigger` (when it
is evaluated: `rootSpanFinished`, or `spanName` with `spanNames` for distributed traces),
`filters` (whether it runs on a trace that fired) and `mode` (`realtime` or `batch`).
Like the UI, this also creates a critical-severity alert, subscribed to your email when
signed in with OAuth.
Explain mutations, irreversible effects, or prerequisites that matter before calling this tool.
Inputs54/100
8 of 13 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
name
Unique within the project, at most 255 characters.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
prompt
LLM instruction describing what to detect in a trace.
Clear enough to supply this argument.
structuredOutput
Payload JSON schema: `{"type": "object", "properties": {"<field>": {"type":
"string"|"number"|"boolean", "description": "...", "enum": [...]}}, "required": [<every
field name>]}`. Field names match `^[a-zA-Z_][a-zA-Z0-9_]*$`.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
sampleRate
Absent or `null` → no sampling (key is not stored).
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
disabled
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › Trigger › oneOf › 0 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › Trigger › oneOf › 1 › spanNames
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › Trigger › oneOf › 1 › type
Not provided
Missing description. Explain what this value means and how to supply it.
filters
Absent = the default filters; `[]` = run on every firing trace. ANDed
`{"column", "operator", "value"}` objects: `total_token_count` (eq, ne, gt, gte, lt,
lte; number), `status` (eq, ne; "error" | "success"), `span_names` and `tags`
(includes, not_includes; list of names).
Clear enough to supply this argument.
mode
Absent or `null` → realtime.
Clear enough to supply this argument.
llmProfileId
Workspace LLM profile id + one of its models. Required together on
self-hosted deployments; rejected on Laminar Cloud.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
model
Not provided
Missing description. Explain what this value means and how to supply it.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "create_signal",
"description": "Create a signal: an LLM judge that runs `prompt` on matching traces and extracts a\npayload matching `structuredOutput`. Three things decide when it runs: `trigger` (when it\nis evaluated: `rootSpanFinished`, or `spanName` with `spanNames` for distributed traces),\n`filters` (whether it runs on a trace that fired) and `mode` (`realtime` or `batch`).\nLike the UI, this also creates a critical-severity alert, subscribed to your email when\nsigned in with OAuth.",
"inputSchema": {
"type": "object",
"$defs": {
"Mode": {
"enum": [
"batch",
"realtime"
],
"type": "string"
},
"Trigger": {
"oneOf": [
{
"type": "object",
"required": [
"type"
],
"properties": {
"type": {
"type": "string",
"const": "rootSpanFinished"
}
},
"description": "The trace's root span finished. Right for most traces."
},
{
"type": "object",
"required": [
"type",
"spanNames"
],
"properties": {
"type": {
"type": "string",
"const": "spanName"
},
"spanNames": {
"type": "array",
"items": {
"type": "string"
}
}
},
"description": "A span with any of these names finished. For distributed traces where no\nsingle span is observably the root."
}
],
"description": "WHEN a signal is evaluated, answerable from a single span batch. A closed set\nrather than a `{column, operator, value}` list: these are the only shapes the\nevaluator implements, so anything else was stored and then never fired."
}
},
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"name",
"prompt",
"structuredOutput"
],
"properties": {
"mode": {
"anyOf": [
{
"$ref": "#/$defs/Mode"
},
{
"type": "null"
}
],
"default": null,
"description": "Absent or `null` → realtime."
},
"name": {
"type": "string",
"description": "Unique within the project, at most 255 characters."
},
"model": {
"type": [
"string",
"null"
],
"default": null
},
"prompt": {
"type": "string",
"description": "LLM instruction describing what to detect in a trace."
},
"filters": {
"type": [
"array",
"null"
],
"items": true,
"default": null,
"description": "Absent = the default filters; `[]` = run on every firing trace. ANDed\n`{\"column\", \"operator\", \"value\"}` objects: `total_token_count` (eq, ne, gt, gte, lt,\nlte; number), `status` (eq, ne; \"error\" | \"success\"), `span_names` and `tags`\n(includes, not_includes; list of names)."
},
"trigger": {
"anyOf": [
{
"$ref": "#/$defs/Trigger"
},
{
"type": "null"
}
],
"default": null,
"description": "Absent or `null` → `rootSpanFinished`."
},
"disabled": {
"type": [
"boolean",
"null"
],
"default": null
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
},
"sampleRate": {
"type": [
"integer",
"null"
],
"format": "int64",
"default": null,
"description": "Absent or `null` → no sampling (key is not stored)."
},
"llmProfileId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Workspace LLM profile id + one of its models. Required together on\nself-hosted deployments; rejected on Laminar Cloud."
},
"structuredOutput": {
"description": "Payload JSON schema: `{\"type\": \"object\", \"properties\": {\"<field>\": {\"type\":\n\"string\"|\"number\"|\"boolean\", \"description\": \"...\", \"enum\": [...]}}, \"required\": [<every\nfield name>]}`. Field names match `^[a-zA-Z_][a-zA-Z0-9_]*$`."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description75/100
Delete an LLM profile. Refused while a signal or feature route uses it.
Explain mutations, irreversible effects, or prerequisites that matter before calling this tool.
Inputs50/100
1 of 2 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
profileId
Not provided
Missing description. Explain what this value means and how to supply it.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "delete_llm_profile",
"description": "Delete an LLM profile. Refused while a signal or feature route uses it.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"profileId"
],
"properties": {
"profileId": {
"type": "string",
"format": "uuid"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description83/100
Delete a signal with its triggers, alerts and events. Irreversible.
The description explains the capability, when to use it, and what to expect.
Inputs50/100
1 of 2 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
signalId
Not provided
Missing description. Explain what this value means and how to supply it.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "delete_signal",
"description": "Delete a signal with its triggers, alerts and events. Irreversible.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"signalId"
],
"properties": {
"signalId": {
"type": "string",
"format": "uuid"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description71/100
Get one LLM profile. Secrets come back masked.
Describe the returned information so the agent knows whether it will answer the request.
Inputs50/100
1 of 2 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
profileId
Not provided
Missing description. Explain what this value means and how to supply it.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_llm_profile",
"description": "Get one LLM profile. Secrets come back masked.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"profileId"
],
"properties": {
"profileId": {
"type": "string",
"format": "uuid"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description70/100
Get one signal with its trigger, filters and mode.
Describe the returned information so the agent knows whether it will answer the request.
Inputs50/100
1 of 2 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
signalId
Not provided
Missing description. Explain what this value means and how to supply it.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_signal",
"description": "Get one signal with its trigger, filters and mode.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"signalId"
],
"properties": {
"signalId": {
"type": "string",
"format": "uuid"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description99/100
Get an LLM-optimized summary of a trace. Returns span hierarchy with timing,
inputs/outputs for LLM spans, and any errors.
Use this when you have a specific trace_id and want to understand what happened.
Use query_laminar_sql first to find trace IDs, then this tool to drill down.
Output includes:
- Span tree with parent-child relationships
- Duration and timing for each span
- Full input/output for LLM spans (truncated for others)
- Exception details if any spans failed
If you need full information that is truncated, you should use query_laminar_sql tool to query spans that you're interested in.
The description explains the capability, when to use it, and what to expect.
Inputs88/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
traceId
The trace ID to retrieve: 32 hex digits, either as a hyphenated GUID (e.g.,
'123e4567-e89b-12d3-a456-426614174000') or unhyphenated as in OpenTelemetry
(e.g., '123e4567e89b12d3a456426614174000')
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "get_trace_context",
"description": "Get an LLM-optimized summary of a trace. Returns span hierarchy with timing,\ninputs/outputs for LLM spans, and any errors.\nUse this when you have a specific trace_id and want to understand what happened.\nUse query_laminar_sql first to find trace IDs, then this tool to drill down.\nOutput includes:\n- Span tree with parent-child relationships\n- Duration and timing for each span\n- Full input/output for LLM spans (truncated for others)\n- Exception details if any spans failed\nIf you need full information that is truncated, you should use query_laminar_sql tool to query spans that you're interested in.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"traceId"
],
"properties": {
"traceId": {
"type": "string",
"pattern": "^(?:[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}|[0-9a-fA-F]{32})$",
"description": "The trace ID to retrieve: 32 hex digits, either as a hyphenated GUID (e.g.,\n'123e4567-e89b-12d3-a456-426614174000') or unhyphenated as in OpenTelemetry\n(e.g., '123e4567e89b12d3a456426614174000')"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description83/100
List the workspace's LLM profiles (provider credentials + models that signals and the
playground run on). Secrets come back masked.
The description explains the capability, when to use it, and what to expect.
Inputs95/100
Input documentation gives an agent enough context to populate the arguments.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_llm_profiles",
"description": "List the workspace's LLM profiles (provider credentials + models that signals and the\nplayground run on). Secrets come back masked.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description85/100
List the Laminar projects this connection can act on, with their workspaces. Pass a
project's `id` as `projectId` to the other tools.
Describe the returned information so the agent knows whether it will answer the request.
Inputs100/100
This tool declares no inputs, so argument descriptions are not needed.
No input fields.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_projects",
"description": "List the Laminar projects this connection can act on, with their workspaces. Pass a\nproject's `id` as `projectId` to the other tools.",
"inputSchema": {
"type": "object",
"properties": {}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description77/100
List a signal's judge-definition (prompt + payload schema) versions, oldest first.
The description explains the capability, when to use it, and what to expect.
Inputs38/100
1 of 2 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
signalId
Not provided
Missing description. Explain what this value means and how to supply it.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_signal_versions",
"description": "List a signal's judge-definition (prompt + payload schema) versions, oldest first.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"signalId"
],
"properties": {
"signalId": {
"type": "string",
"format": "uuid"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description83/100
List the project's signals (LLM analyzers that run on matching traces) with their prompt,
payload schema, trigger, filters and mode. Signal events are queryable with
`query_laminar_sql` (`signal_events`).
The description explains the capability, when to use it, and what to expect.
Inputs75/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
name
Case-insensitive substring of the signal name.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "list_signals",
"description": "List the project's signals (LLM analyzers that run on matching traces) with their prompt,\npayload schema, trigger, filters and mode. Signal events are queryable with\n`query_laminar_sql` (`signal_events`).",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"name": {
"type": [
"string",
"null"
],
"default": null,
"description": "Case-insensitive substring of the signal name."
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description97/100
Execute a read-only SQL query (full ClickHouse SELECT syntax) against this project's Laminar trace data and get rows back as a JSON array. Queries are automatically scoped to your project — never filter on or reference a `project_id` column. Only SELECT is allowed.
<tables>
TABLE spans — Individual spans within traces: timing, tokens, costs, and LLM-specific data.
span_id UUID — Unique id of the span
trace_id UUID — Id of the trace this span belongs to
parent_span_id UUID — Id of the parent span
name String — Name of the span
path String — Hierarchical path, e.g. 'outer.inner'
span_type String (enum span_type) — Span type, e.g. 'LLM', 'TOOL', 'DEFAULT'
status String (enum status) — 'success' or 'error'
start_time DateTime64(9,'UTC') — When the span started
end_time DateTime64(9,'UTC') — When the span ended
duration Decimal(18,9) — Duration in seconds
input String — Span input as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
output String — Span output as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
request_model String — LLM model requested
response_model String — LLM model in the response
model String — coalesce(request_model, response_model)
provider String — LLM provider, e.g. 'openai', 'anthropic'
input_tokens Int64 — Input tokens
output_tokens Int64 — Output tokens
total_tokens Int64 — Total tokens
cache_read_input_tokens UInt64 — Tokens read from prompt cache
cache_creation_input_tokens UInt64 — Tokens written to prompt cache
reasoning_tokens UInt64 — Reasoning tokens
input_cost Float64 — Input cost
output_cost Float64 — Output cost
total_cost Float64 — Total cost
attributes String — Span attributes as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
tags Array(String) — Span-level tags
tool_definitions String — Tool definitions exposed to the LLM span as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
events Array(Tuple(timestamp Int64, name String, attributes String)) — Span events; timestamp is unix nanoseconds, attributes is stringified JSON
TABLE traces — Top-level trace records aggregating span data with session and user context. Filter on start_time/end_time to bound the scan.
id UUID — Unique id of the trace
trace_type String (enum trace_type) — 'DEFAULT', 'EVALUATION', or 'PLAYGROUND'
metadata String — Trace metadata as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
start_time DateTime64(9,'UTC') — When the trace started
end_time DateTime64(9,'UTC') — When the trace ended
duration Float64 — Duration in seconds
input_tokens Int64 — Input tokens
output_tokens Int64 — Output tokens
total_tokens Int64 — Total tokens
cache_read_input_tokens UInt64 — Tokens read from prompt cache
cache_creation_input_tokens UInt64 — Tokens written to prompt cache
reasoning_tokens UInt64 — Reasoning tokens
input_cost Float64 — Input cost
output_cost Float64 — Output cost
total_cost Float64 — Total cost
status String (enum status) — 'error' if any span errored, else 'success'
user_id String — User id sent with the trace
session_id String — Session identifier
top_span_id UUID — Id of the top-level span
top_span_name String — Name of the top-level span
top_span_type String (enum span_type) — Type of the top-level span
tags Array(String) — Union of span-level tags in the trace
trace_tags Array(String) — Tags applied directly to the trace
span_names Array(String) — De-duplicated span names in the trace
agent_input String — Extracted agent task / user input for the trace (stringified JSON / raw string) — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
has_browser_session Bool — Whether the trace has a recorded browser session
signal_events Array(Tuple(event_id UUID, signal_id UUID, severity UInt8, payload String, timestamp DateTime64(9, 'UTC'), updated_at DateTime64(9, 'UTC'))) — Signal events that fired on this trace, read from signal_events when the query runs, so discarded events and those of deleted signals are excluded. ARRAY JOIN signal_events AS e to unnest, then read e.payload. Reading this column scans the project's signal events created since the query's earliest start_time, so narrow by time first. e.timestamp is when the event fired, e.updated_at when it last changed; signal names live in the signals table, join on signal_id. e.payload is the full event JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
clusters Array(Tuple(id UUID, signal_id UUID, name String, level UInt8, parent_id UUID, num_signal_events UInt32, created_at DateTime64(9), updated_at DateTime64(9))) — Named clusters (L1 and ancestors) this trace's signal events belong to. Empty until events are clustered. Prefer this over joining signal_events to clusters. Ancestors are already included as their own elements; walk between them with parent_id. num_signal_events counts the whole cluster, not this trace
agent_versions Array(Tuple(agent_hash String, version_hash String, model String)) — Agent versions that ran in this trace, one element per (agent, version, model) seen; model is '' when unknown. ARRAY JOIN agent_versions AS v to compare error rates or signal events across versions and models, counting traces with uniqExact(id), not count(), since a trace repeats once per model. Beta: shape may change. Traces are not tagged instantly, so recent traces may be empty
release String — Release of the code that produced the trace, as sent by the SDK (usually the short git SHA, 7 lowercase hex chars). Empty when the SDK sent none
TABLE trace_outputs — Extracted final agent output per trace: the output-message array of the last LLM call on the trace's main-agent path.
trace_id UUID — Id of the trace
start_time DateTime64(9,'UTC') — Start time of the trace this output belongs to
agent_output Array(String) — Output messages of the trace's final LLM call, one stringified JSON message per element
TABLE evaluation_trials — Results from evaluations, one row per trial: scores, executor output, and denormalized trace data. `evaluation_datapoints` is an alias for this table.
id UUID — Unique id of the evaluation datapoint
evaluation_id UUID — Id of the evaluation
evaluation_name String — Name of the evaluation
trace_id UUID — Id of the associated trace
created_at DateTime64(9,'UTC') — When the Trial was first saved
data String — Input data — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
target String — Target / expected output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
metadata String — Metadata as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
executor_output String — Executor output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
index UInt64 — Index within the evaluation
task_id String — Task (test case) id; rows sharing a task_id within an evaluation are repeated trials of one task
group_id String — Group id of the evaluation run
scores String — Stringified JSON object: score name -> value — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
dataset_id UUID — Linked dataset id (nil if none)
dataset_datapoint_id UUID — Linked dataset datapoint id (nil if none)
dataset_datapoint_created_at DateTime64(9,'UTC') — When the linked dataset datapoint was created (unix epoch if none)
updated_at DateTime64(9,'UTC') — When the Trial was last updated
duration Decimal(18,9) — Duration in seconds from the associated trace
start_time DateTime64(9,'UTC') — Start time of the associated trace
end_time DateTime64(9,'UTC') — End time of the associated trace
input_cost Float64 — Input cost from the trace
output_cost Float64 — Output cost from the trace
total_cost Float64 — Total cost from the associated trace
input_tokens Int64 — Input tokens from the trace
output_tokens Int64 — Output tokens from the trace
total_tokens Int64 — Total tokens from the associated trace
trace_status LowCardinality(String) (enum status) — Status of the associated trace
trace_metadata String — Metadata of the associated trace as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
trace_tags Array(String) — Tags of the associated trace
top_span_id UUID — Id of the top-level span of the associated trace
trace_spans Array(Tuple(name String, duration Float64, type String)) — Spans of the associated trace; `type` is the numeric span type as a string
TABLE dataset_datapoints — Data points in datasets with input data, targets, and metadata.
id UUID — Unique id
created_at DateTime64(9,'UTC') — When created
dataset_id UUID — Id of the dataset
dataset_name String — Name of the dataset
data String — Input data — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
target String — Target / expected output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
metadata String — Additional metadata — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
task_id String — Task id evaluations group trials by; unique within the dataset
TABLE dataset_datapoint_versions — Same columns as dataset_datapoints, but every version of each datapoint rather than only the latest.
id UUID — Unique id
created_at DateTime64(9,'UTC') — When created
dataset_id UUID — Id of the dataset
dataset_name String — Name of the dataset
data String — Input data — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
target String — Target / expected output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
metadata String — Additional metadata — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
task_id String — Task id evaluations group trials by; unique within the dataset
TABLE logs — Log entries with severity, body, and trace correlation.
log_id UUID — Unique id of the log
time DateTime64(9,'UTC') — When the log occurred
observed_time DateTime64(9,'UTC') — When the log was observed
severity_number UInt8 — Numeric severity
severity_text String — Severity text
body String — Log body — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
attributes String — Attributes as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
trace_id UUID — Id of the trace
span_id UUID — Id of the span
flags UInt32 — Flags for the log
event_name String — Event name
TABLE signal_runs — Execution records for signals with status and error info.
signal_id UUID — Id of the signal
job_id UUID — Id of the job
trigger_id UUID — Id of the trigger
run_id UUID — Id of the run
trace_id UUID — Id of the trace
status String (enum signal_run_status) — Pipeline stage: 'PENDING' (elected, waiting on the agent), 'PROCESSING' (agent running), 'COMPLETED', 'FAILED'
mode String (enum signal_run_mode) — 'BATCH' (historical backfill) or 'REALTIME'
event_id UUID — Id of the produced event
error_message String — Error message if the run failed
updated_at DateTime64(9,'UTC') — When last updated
input_tokens UInt32 — Input tokens spent evaluating the signal
cache_read_tokens UInt32 — Cached input tokens read (a subset of input_tokens)
output_tokens UInt32 — Output tokens produced by the signal agent
signal_version UInt32 — Signal definition version the run used; 0 predates versioning
TABLE signal_events — Events emitted by signals during execution. Excludes L0 clusters in `clusters`.
id UUID — Unique id of the signal event
signal_id UUID — Id of the signal
signal_name String — Current name of the signal
trace_id UUID — Id of the trace
run_id UUID — Id of the run
name String — Alias of signal_name, kept for older queries; use signal_name
payload String — Payload as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
timestamp DateTime64(9,'UTC') — When the event occurred
severity UInt8 — 0 = INFO, 1 = WARNING, 2 = CRITICAL
clusters Array(UUID) — Cluster ids this event belongs to (excludes L0)
leaf_clusters Array(UUID) — L1 (finest named) cluster ids this event belongs to
cluster_details Array(Tuple(id UUID, name String, level UInt8)) — Named clusters this event belongs to, with name and level (excludes L0)
signal_version UInt32 — Signal definition version that produced the event; 0 predates versioning
TABLE clusters — Hierarchy of clusters of similar signal events. Excludes L0 clusters. Use this when the user asks about clusters / clustering / grouped signal events.
id UUID — Unique id of the cluster
signal_id UUID — Id of the signal the cluster belongs to
name String — Human-readable cluster name
level UInt8 — Level in the hierarchy; higher = coarser
parent_id UUID — Parent cluster id (nil UUID for top-level)
num_signal_events UInt32 — Number of clustered event summaries in the cluster (an event contributes once per summary)
num_children_clusters UInt16 — Number of immediate child clusters
created_at DateTime64(9,'UTC') — When created
updated_at DateTime64(9,'UTC') — When last updated
TABLE signals — Signal definitions stored in Postgres: prompt, output schema, model, and current version. Join to signal_events / signal_runs / traces.signal_events on signals.id = signal_id; for just the name, use signal_events.signal_name.
id UUID — Unique id of the signal
name String — Signal name, unique per project
prompt String — Signal definition prompt
structured_output_schema String — JSON schema the signal's output conforms to
metadata String — Signal metadata, as a JSON object
llm_model Nullable(String) — Model the signal runs on; NULL when it uses the deployment default
created_at DateTime64(6) — When the signal was created
version Int32 — Current definition version
TABLE evaluations — Evaluations stored in Postgres: name, group and metadata. Join to evaluation_trials on evaluations.id = evaluation_id; for just the name, use evaluation_trials.evaluation_name.
id UUID — Unique id of the evaluation
name String — Evaluation name
group_id String — Group the evaluation belongs to
metadata Nullable(String) — Evaluation metadata, as a JSON object
created_at DateTime64(6) — When the evaluation was created
TABLE datasets — Datasets stored in Postgres. Join to dataset_datapoints on datasets.id = dataset_id; for just the name, use dataset_datapoints.dataset_name.
id UUID — Unique id of the dataset
name String — Dataset name
indexed_on Nullable(String) — Legacy indexing setting, unused by Laminar; usually NULL
created_at DateTime64(6) — When the dataset was created
TABLE dashboards — Project dashboards stored in Postgres. Join to dashboard_charts on dashboards.id = dashboard_id.
id UUID — Unique id of the dashboard
name String — Dashboard name
created_at DateTime64(6) — When the dashboard was created
TABLE dashboard_charts — Charts on project dashboards, stored in Postgres. The grid is 12 columns wide; a chart's position is settings.layout {x, y, w, h} in grid units.
id UUID — Unique id of the chart
dashboard_id UUID — Id of the dashboard the chart is on
name String — Chart title
query String — SQL the chart runs, with {start_time:DateTime64} / {end_time:DateTime64} / {interval_unit:String} parameters
settings String — Chart settings as a JSON object: {config, layout, queryStructure}. Read fields with JSONExtract, e.g. JSONExtractInt(settings, 'layout', 'y')
created_at DateTime64(6) — When the chart was created
TABLE alerts — Signal alerts stored in Postgres. Join to alert_targets and alert_filters on alerts.id = alert_id, and to signals on alerts.signal_id = signals.id.
id UUID — Unique id of the alert
name String — Alert name
type String — SIGNAL_EVENT (notify on each matching signal event) or NEW_CLUSTER (notify when a new cluster forms)
signal_id UUID — Id of the signal the alert watches
metadata String — Alert settings as a JSON object: severities (levels 0 info, 1 warning, 2 critical; absent means [2]), skipSimilar and disabled (true when set, absent otherwise)
created_at DateTime64(6) — When the alert was created
TABLE alert_targets — Where alerts are delivered, stored in Postgres: one row per Slack channel or subscribed member email.
id UUID — Unique id of the target
alert_id UUID — Id of the alert
type String — SLACK or EMAIL
channel_id Nullable(String) — Slack channel id; NULL for EMAIL targets
channel_name Nullable(String) — Slack channel name; NULL for EMAIL targets
email Nullable(String) — Subscribed member's email; NULL for SLACK targets
created_at DateTime64(6) — When the target was added
TABLE alert_filters — Conditions an alert's signal events must match, stored in Postgres. Conditions within one row are ANDed; an alert fires if any of its rows matches (or always, with no rows).
id UUID — Unique id of the filter group
alert_id UUID — Id of the alert
value String — JSON array of conditions {column, operator, value}; column is a field of the signal's output
created_at DateTime64(6) — When the filter group was added
TABLE table_views — Saved views of the UI's data tables, stored in Postgres and shared by the project. A view is its table's column layout plus filters, search and sort.
id UUID — Unique id of the view
resource String — Table the view belongs to: traces, spans, sessions, evaluations, evaluation-v1.1 (an evaluation's datapoints), datasets, dataset (a dataset's datapoints), queues, playgrounds, debugger-sessions, signal-runs, or signal-events:<signal id>
name String — View name, unique per resource
config String — View as a JSON object: columnOrder (column ids), columnVisibility ({id: bool}, every column), columnSizing ({id: px}), customColumns ([{name, sql, dataType}]), filters ([{column, operator, value}]), search, sortBy, sortDirection (asc/desc), and isDefault (true on the one view per resource teammates land on). Custom column ids are custom:<name>
created_at DateTime64(6) — When the view was created
updated_at DateTime64(6) — When the view was last saved
TABLE labeling_queue_items — Per-item rows of labeling queues.
id UUID — Unique id of the queue item
queue_id UUID — Id of the labeling queue
payload String — Original seeded {data, target, metadata} as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
metadata String — Additional metadata — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
status UInt8 — 0 = unlabeled, 1 = approved
edit String — Current canonical target as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped
created_at DateTime64(3,'UTC') — When created
updated_at DateTime64(3,'UTC') — When last updated
</tables>
<enums>
span_type: 'DEFAULT', 'LLM', 'EXECUTOR', 'EVALUATOR', 'EVALUATION', 'TOOL', 'HUMAN_EVALUATOR', 'CACHED', 'UNKNOWN'
trace_type: 'DEFAULT', 'EVALUATION', 'PLAYGROUND'
status: 'success', 'error'
signal_run_status: 'PENDING', 'PROCESSING', 'COMPLETED', 'FAILED', 'UNKNOWN'
signal_run_mode: 'BATCH', 'REALTIME', 'UNKNOWN'
</enums>
<joins>
- spans.trace_id = traces.id
- signal_events.trace_id = traces.id
- trace_outputs.trace_id = traces.id
- has(signal_events.clusters, clusters.id) to match events to the specific clusters they belong to
(clusters.signal_id = signal_events.signal_id only scopes by signal — it is a many-to-many cross
product, NOT an event-to-cluster match).
- signals.id = signal_events.signal_id (and signal_runs.signal_id, e.signal_id after ARRAY JOIN traces.signal_events AS e)
for a signal's prompt, schema, model or version. Likewise evaluations.id =
evaluation_trials.evaluation_id and datasets.id = dataset_datapoints.dataset_id. Names need no
join: signal_events.signal_name, evaluation_trials.evaluation_name, dataset_datapoints.dataset_name.
- For per-trace signal/cluster questions prefer traces.signal_events and traces.clusters over joining
signal_events. ARRAY JOIN signal_events AS e, ARRAY JOIN clusters AS c.
- Top-level clusters have parent_id = the nil UUID '00000000-0000-0000-0000-000000000000' (NOT SQL
NULL): filter with parent_id = toUUID('00000000-0000-0000-0000-000000000000'), not IS NULL.
</joins>
<parameters>
Use the `parameters` argument for {name:Type} placeholders in the query, e.g. {trace_id:UUID}.
</parameters>
<examples>
- Recent traces: SELECT id, start_time, total_cost FROM traces ORDER BY start_time DESC LIMIT 10
- LLM spans: SELECT name, model, input, output FROM spans WHERE span_type = 'LLM' LIMIT 20
- Errors: SELECT trace_id, name, status FROM spans WHERE status = 'error' LIMIT 20
- Top clusters: SELECT name, num_signal_events FROM clusters ORDER BY num_signal_events DESC LIMIT 10
</examples>
The description explains the capability, when to use it, and what to expect.
Inputs50/100
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
query
ClickHouse SQL query. Must be SELECT only. See the `query_laminar_sql` tool description for the
full table/column schema, enums, joins, and example queries.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
parameters
Query parameters for {name:Type} placeholders, e.g., {trace_id:UUID}
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Resolve the mismatch between this description and its schema constraints.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "query_laminar_sql",
"description": "Execute a read-only SQL query (full ClickHouse SELECT syntax) against this project's Laminar trace data and get rows back as a JSON array. Queries are automatically scoped to your project — never filter on or reference a `project_id` column. Only SELECT is allowed.\n\n<tables>\nTABLE spans — Individual spans within traces: timing, tokens, costs, and LLM-specific data.\n span_id UUID — Unique id of the span\n trace_id UUID — Id of the trace this span belongs to\n parent_span_id UUID — Id of the parent span\n name String — Name of the span\n path String — Hierarchical path, e.g. 'outer.inner'\n span_type String (enum span_type) — Span type, e.g. 'LLM', 'TOOL', 'DEFAULT'\n status String (enum status) — 'success' or 'error'\n start_time DateTime64(9,'UTC') — When the span started\n end_time DateTime64(9,'UTC') — When the span ended\n duration Decimal(18,9) — Duration in seconds\n input String — Span input as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n output String — Span output as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n request_model String — LLM model requested\n response_model String — LLM model in the response\n model String — coalesce(request_model, response_model)\n provider String — LLM provider, e.g. 'openai', 'anthropic'\n input_tokens Int64 — Input tokens\n output_tokens Int64 — Output tokens\n total_tokens Int64 — Total tokens\n cache_read_input_tokens UInt64 — Tokens read from prompt cache\n cache_creation_input_tokens UInt64 — Tokens written to prompt cache\n reasoning_tokens UInt64 — Reasoning tokens\n input_cost Float64 — Input cost\n output_cost Float64 — Output cost\n total_cost Float64 — Total cost\n attributes String — Span attributes as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n tags Array(String) — Span-level tags\n tool_definitions String — Tool definitions exposed to the LLM span as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n events Array(Tuple(timestamp Int64, name String, attributes String)) — Span events; timestamp is unix nanoseconds, attributes is stringified JSON\n\nTABLE traces — Top-level trace records aggregating span data with session and user context. Filter on start_time/end_time to bound the scan.\n id UUID — Unique id of the trace\n trace_type String (enum trace_type) — 'DEFAULT', 'EVALUATION', or 'PLAYGROUND'\n metadata String — Trace metadata as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n start_time DateTime64(9,'UTC') — When the trace started\n end_time DateTime64(9,'UTC') — When the trace ended\n duration Float64 — Duration in seconds\n input_tokens Int64 — Input tokens\n output_tokens Int64 — Output tokens\n total_tokens Int64 — Total tokens\n cache_read_input_tokens UInt64 — Tokens read from prompt cache\n cache_creation_input_tokens UInt64 — Tokens written to prompt cache\n reasoning_tokens UInt64 — Reasoning tokens\n input_cost Float64 — Input cost\n output_cost Float64 — Output cost\n total_cost Float64 — Total cost\n status String (enum status) — 'error' if any span errored, else 'success'\n user_id String — User id sent with the trace\n session_id String — Session identifier\n top_span_id UUID — Id of the top-level span\n top_span_name String — Name of the top-level span\n top_span_type String (enum span_type) — Type of the top-level span\n tags Array(String) — Union of span-level tags in the trace\n trace_tags Array(String) — Tags applied directly to the trace\n span_names Array(String) — De-duplicated span names in the trace\n agent_input String — Extracted agent task / user input for the trace (stringified JSON / raw string) — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n has_browser_session Bool — Whether the trace has a recorded browser session\n signal_events Array(Tuple(event_id UUID, signal_id UUID, severity UInt8, payload String, timestamp DateTime64(9, 'UTC'), updated_at DateTime64(9, 'UTC'))) — Signal events that fired on this trace, read from signal_events when the query runs, so discarded events and those of deleted signals are excluded. ARRAY JOIN signal_events AS e to unnest, then read e.payload. Reading this column scans the project's signal events created since the query's earliest start_time, so narrow by time first. e.timestamp is when the event fired, e.updated_at when it last changed; signal names live in the signals table, join on signal_id. e.payload is the full event JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n clusters Array(Tuple(id UUID, signal_id UUID, name String, level UInt8, parent_id UUID, num_signal_events UInt32, created_at DateTime64(9), updated_at DateTime64(9))) — Named clusters (L1 and ancestors) this trace's signal events belong to. Empty until events are clustered. Prefer this over joining signal_events to clusters. Ancestors are already included as their own elements; walk between them with parent_id. num_signal_events counts the whole cluster, not this trace\n agent_versions Array(Tuple(agent_hash String, version_hash String, model String)) — Agent versions that ran in this trace, one element per (agent, version, model) seen; model is '' when unknown. ARRAY JOIN agent_versions AS v to compare error rates or signal events across versions and models, counting traces with uniqExact(id), not count(), since a trace repeats once per model. Beta: shape may change. Traces are not tagged instantly, so recent traces may be empty\n release String — Release of the code that produced the trace, as sent by the SDK (usually the short git SHA, 7 lowercase hex chars). Empty when the SDK sent none\n\nTABLE trace_outputs — Extracted final agent output per trace: the output-message array of the last LLM call on the trace's main-agent path.\n trace_id UUID — Id of the trace\n start_time DateTime64(9,'UTC') — Start time of the trace this output belongs to\n agent_output Array(String) — Output messages of the trace's final LLM call, one stringified JSON message per element\n\nTABLE evaluation_trials — Results from evaluations, one row per trial: scores, executor output, and denormalized trace data. `evaluation_datapoints` is an alias for this table.\n id UUID — Unique id of the evaluation datapoint\n evaluation_id UUID — Id of the evaluation\n evaluation_name String — Name of the evaluation\n trace_id UUID — Id of the associated trace\n created_at DateTime64(9,'UTC') — When the Trial was first saved\n data String — Input data — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n target String — Target / expected output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n metadata String — Metadata as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n executor_output String — Executor output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n index UInt64 — Index within the evaluation\n task_id String — Task (test case) id; rows sharing a task_id within an evaluation are repeated trials of one task\n group_id String — Group id of the evaluation run\n scores String — Stringified JSON object: score name -> value — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n dataset_id UUID — Linked dataset id (nil if none)\n dataset_datapoint_id UUID — Linked dataset datapoint id (nil if none)\n dataset_datapoint_created_at DateTime64(9,'UTC') — When the linked dataset datapoint was created (unix epoch if none)\n updated_at DateTime64(9,'UTC') — When the Trial was last updated\n duration Decimal(18,9) — Duration in seconds from the associated trace\n start_time DateTime64(9,'UTC') — Start time of the associated trace\n end_time DateTime64(9,'UTC') — End time of the associated trace\n input_cost Float64 — Input cost from the trace\n output_cost Float64 — Output cost from the trace\n total_cost Float64 — Total cost from the associated trace\n input_tokens Int64 — Input tokens from the trace\n output_tokens Int64 — Output tokens from the trace\n total_tokens Int64 — Total tokens from the associated trace\n trace_status LowCardinality(String) (enum status) — Status of the associated trace\n trace_metadata String — Metadata of the associated trace as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n trace_tags Array(String) — Tags of the associated trace\n top_span_id UUID — Id of the top-level span of the associated trace\n trace_spans Array(Tuple(name String, duration Float64, type String)) — Spans of the associated trace; `type` is the numeric span type as a string\n\nTABLE dataset_datapoints — Data points in datasets with input data, targets, and metadata.\n id UUID — Unique id\n created_at DateTime64(9,'UTC') — When created\n dataset_id UUID — Id of the dataset\n dataset_name String — Name of the dataset\n data String — Input data — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n target String — Target / expected output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n metadata String — Additional metadata — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n task_id String — Task id evaluations group trials by; unique within the dataset\n\nTABLE dataset_datapoint_versions — Same columns as dataset_datapoints, but every version of each datapoint rather than only the latest.\n id UUID — Unique id\n created_at DateTime64(9,'UTC') — When created\n dataset_id UUID — Id of the dataset\n dataset_name String — Name of the dataset\n data String — Input data — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n target String — Target / expected output — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n metadata String — Additional metadata — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n task_id String — Task id evaluations group trials by; unique within the dataset\n\nTABLE logs — Log entries with severity, body, and trace correlation.\n log_id UUID — Unique id of the log\n time DateTime64(9,'UTC') — When the log occurred\n observed_time DateTime64(9,'UTC') — When the log was observed\n severity_number UInt8 — Numeric severity\n severity_text String — Severity text\n body String — Log body — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n attributes String — Attributes as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n trace_id UUID — Id of the trace\n span_id UUID — Id of the span\n flags UInt32 — Flags for the log\n event_name String — Event name\n\nTABLE signal_runs — Execution records for signals with status and error info.\n signal_id UUID — Id of the signal\n job_id UUID — Id of the job\n trigger_id UUID — Id of the trigger\n run_id UUID — Id of the run\n trace_id UUID — Id of the trace\n status String (enum signal_run_status) — Pipeline stage: 'PENDING' (elected, waiting on the agent), 'PROCESSING' (agent running), 'COMPLETED', 'FAILED'\n mode String (enum signal_run_mode) — 'BATCH' (historical backfill) or 'REALTIME'\n event_id UUID — Id of the produced event\n error_message String — Error message if the run failed\n updated_at DateTime64(9,'UTC') — When last updated\n input_tokens UInt32 — Input tokens spent evaluating the signal\n cache_read_tokens UInt32 — Cached input tokens read (a subset of input_tokens)\n output_tokens UInt32 — Output tokens produced by the signal agent\n signal_version UInt32 — Signal definition version the run used; 0 predates versioning\n\nTABLE signal_events — Events emitted by signals during execution. Excludes L0 clusters in `clusters`.\n id UUID — Unique id of the signal event\n signal_id UUID — Id of the signal\n signal_name String — Current name of the signal\n trace_id UUID — Id of the trace\n run_id UUID — Id of the run\n name String — Alias of signal_name, kept for older queries; use signal_name\n payload String — Payload as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n timestamp DateTime64(9,'UTC') — When the event occurred\n severity UInt8 — 0 = INFO, 1 = WARNING, 2 = CRITICAL\n clusters Array(UUID) — Cluster ids this event belongs to (excludes L0)\n leaf_clusters Array(UUID) — L1 (finest named) cluster ids this event belongs to\n cluster_details Array(Tuple(id UUID, name String, level UInt8)) — Named clusters this event belongs to, with name and level (excludes L0)\n signal_version UInt32 — Signal definition version that produced the event; 0 predates versioning\n\nTABLE clusters — Hierarchy of clusters of similar signal events. Excludes L0 clusters. Use this when the user asks about clusters / clustering / grouped signal events.\n id UUID — Unique id of the cluster\n signal_id UUID — Id of the signal the cluster belongs to\n name String — Human-readable cluster name\n level UInt8 — Level in the hierarchy; higher = coarser\n parent_id UUID — Parent cluster id (nil UUID for top-level)\n num_signal_events UInt32 — Number of clustered event summaries in the cluster (an event contributes once per summary)\n num_children_clusters UInt16 — Number of immediate child clusters\n created_at DateTime64(9,'UTC') — When created\n updated_at DateTime64(9,'UTC') — When last updated\n\nTABLE signals — Signal definitions stored in Postgres: prompt, output schema, model, and current version. Join to signal_events / signal_runs / traces.signal_events on signals.id = signal_id; for just the name, use signal_events.signal_name.\n id UUID — Unique id of the signal\n name String — Signal name, unique per project\n prompt String — Signal definition prompt\n structured_output_schema String — JSON schema the signal's output conforms to\n metadata String — Signal metadata, as a JSON object\n llm_model Nullable(String) — Model the signal runs on; NULL when it uses the deployment default\n created_at DateTime64(6) — When the signal was created\n version Int32 — Current definition version\n\nTABLE evaluations — Evaluations stored in Postgres: name, group and metadata. Join to evaluation_trials on evaluations.id = evaluation_id; for just the name, use evaluation_trials.evaluation_name.\n id UUID — Unique id of the evaluation\n name String — Evaluation name\n group_id String — Group the evaluation belongs to\n metadata Nullable(String) — Evaluation metadata, as a JSON object\n created_at DateTime64(6) — When the evaluation was created\n\nTABLE datasets — Datasets stored in Postgres. Join to dataset_datapoints on datasets.id = dataset_id; for just the name, use dataset_datapoints.dataset_name.\n id UUID — Unique id of the dataset\n name String — Dataset name\n indexed_on Nullable(String) — Legacy indexing setting, unused by Laminar; usually NULL\n created_at DateTime64(6) — When the dataset was created\n\nTABLE dashboards — Project dashboards stored in Postgres. Join to dashboard_charts on dashboards.id = dashboard_id.\n id UUID — Unique id of the dashboard\n name String — Dashboard name\n created_at DateTime64(6) — When the dashboard was created\n\nTABLE dashboard_charts — Charts on project dashboards, stored in Postgres. The grid is 12 columns wide; a chart's position is settings.layout {x, y, w, h} in grid units.\n id UUID — Unique id of the chart\n dashboard_id UUID — Id of the dashboard the chart is on\n name String — Chart title\n query String — SQL the chart runs, with {start_time:DateTime64} / {end_time:DateTime64} / {interval_unit:String} parameters\n settings String — Chart settings as a JSON object: {config, layout, queryStructure}. Read fields with JSONExtract, e.g. JSONExtractInt(settings, 'layout', 'y')\n created_at DateTime64(6) — When the chart was created\n\nTABLE alerts — Signal alerts stored in Postgres. Join to alert_targets and alert_filters on alerts.id = alert_id, and to signals on alerts.signal_id = signals.id.\n id UUID — Unique id of the alert\n name String — Alert name\n type String — SIGNAL_EVENT (notify on each matching signal event) or NEW_CLUSTER (notify when a new cluster forms)\n signal_id UUID — Id of the signal the alert watches\n metadata String — Alert settings as a JSON object: severities (levels 0 info, 1 warning, 2 critical; absent means [2]), skipSimilar and disabled (true when set, absent otherwise)\n created_at DateTime64(6) — When the alert was created\n\nTABLE alert_targets — Where alerts are delivered, stored in Postgres: one row per Slack channel or subscribed member email.\n id UUID — Unique id of the target\n alert_id UUID — Id of the alert\n type String — SLACK or EMAIL\n channel_id Nullable(String) — Slack channel id; NULL for EMAIL targets\n channel_name Nullable(String) — Slack channel name; NULL for EMAIL targets\n email Nullable(String) — Subscribed member's email; NULL for SLACK targets\n created_at DateTime64(6) — When the target was added\n\nTABLE alert_filters — Conditions an alert's signal events must match, stored in Postgres. Conditions within one row are ANDed; an alert fires if any of its rows matches (or always, with no rows).\n id UUID — Unique id of the filter group\n alert_id UUID — Id of the alert\n value String — JSON array of conditions {column, operator, value}; column is a field of the signal's output\n created_at DateTime64(6) — When the filter group was added\n\nTABLE table_views — Saved views of the UI's data tables, stored in Postgres and shared by the project. A view is its table's column layout plus filters, search and sort.\n id UUID — Unique id of the view\n resource String — Table the view belongs to: traces, spans, sessions, evaluations, evaluation-v1.1 (an evaluation's datapoints), datasets, dataset (a dataset's datapoints), queues, playgrounds, debugger-sessions, signal-runs, or signal-events:<signal id>\n name String — View name, unique per resource\n config String — View as a JSON object: columnOrder (column ids), columnVisibility ({id: bool}, every column), columnSizing ({id: px}), customColumns ([{name, sql, dataType}]), filters ([{column, operator, value}]), search, sortBy, sortDirection (asc/desc), and isDefault (true on the one view per resource teammates land on). Custom column ids are custom:<name>\n created_at DateTime64(6) — When the view was created\n updated_at DateTime64(6) — When the view was last saved\n\nTABLE labeling_queue_items — Per-item rows of labeling queues.\n id UUID — Unique id of the queue item\n queue_id UUID — Id of the labeling queue\n payload String — Original seeded {data, target, metadata} as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n metadata String — Additional metadata — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n status UInt8 — 0 = unlabeled, 1 = approved\n edit String — Current canonical target as stringified JSON — LARGE: select as substring(col, 1, 2000), never raw, or the row is dropped\n created_at DateTime64(3,'UTC') — When created\n updated_at DateTime64(3,'UTC') — When last updated\n\n</tables>\n<enums>\nspan_type: 'DEFAULT', 'LLM', 'EXECUTOR', 'EVALUATOR', 'EVALUATION', 'TOOL', 'HUMAN_EVALUATOR', 'CACHED', 'UNKNOWN'\ntrace_type: 'DEFAULT', 'EVALUATION', 'PLAYGROUND'\nstatus: 'success', 'error'\nsignal_run_status: 'PENDING', 'PROCESSING', 'COMPLETED', 'FAILED', 'UNKNOWN'\nsignal_run_mode: 'BATCH', 'REALTIME', 'UNKNOWN'\n</enums>\n\n<joins>\n- spans.trace_id = traces.id\n- signal_events.trace_id = traces.id\n- trace_outputs.trace_id = traces.id\n- has(signal_events.clusters, clusters.id) to match events to the specific clusters they belong to\n (clusters.signal_id = signal_events.signal_id only scopes by signal — it is a many-to-many cross\n product, NOT an event-to-cluster match).\n- signals.id = signal_events.signal_id (and signal_runs.signal_id, e.signal_id after ARRAY JOIN traces.signal_events AS e)\n for a signal's prompt, schema, model or version. Likewise evaluations.id =\n evaluation_trials.evaluation_id and datasets.id = dataset_datapoints.dataset_id. Names need no\n join: signal_events.signal_name, evaluation_trials.evaluation_name, dataset_datapoints.dataset_name.\n- For per-trace signal/cluster questions prefer traces.signal_events and traces.clusters over joining\n signal_events. ARRAY JOIN signal_events AS e, ARRAY JOIN clusters AS c.\n- Top-level clusters have parent_id = the nil UUID '00000000-0000-0000-0000-000000000000' (NOT SQL\n NULL): filter with parent_id = toUUID('00000000-0000-0000-0000-000000000000'), not IS NULL.\n</joins>\n\n<parameters>\nUse the `parameters` argument for {name:Type} placeholders in the query, e.g. {trace_id:UUID}.\n</parameters>\n\n<examples>\n- Recent traces: SELECT id, start_time, total_cost FROM traces ORDER BY start_time DESC LIMIT 10\n- LLM spans: SELECT name, model, input, output FROM spans WHERE span_type = 'LLM' LIMIT 20\n- Errors: SELECT trace_id, name, status FROM spans WHERE status = 'error' LIMIT 20\n- Top clusters: SELECT name, num_signal_events FROM clusters ORDER BY num_signal_events DESC LIMIT 10\n</examples>",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"query"
],
"properties": {
"query": {
"type": "string",
"description": "ClickHouse SQL query. Must be SELECT only. See the `query_laminar_sql` tool description for the\nfull table/column schema, enums, joins, and example queries."
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
},
"parameters": {
"type": "object",
"default": {},
"description": "Query parameters for {name:Type} placeholders, e.g., {trace_id:UUID}",
"additionalProperties": true
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description68/100
Update an LLM profile; only the fields you pass change.
Explain mutations, irreversible effects, or prerequisites that matter before calling this tool.
Inputs25/100
5 of 18 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
profileId
Not provided
Missing description. Explain what this value means and how to supply it.
name
Not provided
Missing description. Explain what this value means and how to supply it.
provider
Changing it requires `config`, which then starts empty instead of from the stored one.
Clear enough to supply this argument.
$defs › ProfileAuth › oneOf › 0 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileAuth › oneOf › 1 › accessKeyId
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileAuth › oneOf › 1 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileAuth › oneOf › 2 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfigPatch › region
`""` clears it; likewise `resourceId`, `baseUrl` and `apiVersion`.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
$defs › ProfileConfigPatch › resourceId
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfigPatch › baseUrl
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfigPatch › apiVersion
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileConfigPatch › headerNames
Custom header names, with their values in `secrets.headers`. `[]` removes every header
and its stored value.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
$defs › ProfileSecrets › apiKey
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileSecrets › secretAccessKey
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileSecrets › token
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › ProfileSecrets › headers
Not provided
Missing description. Explain what this value means and how to supply it.
models
Replaces the whole list.
Clear enough to supply this argument.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "update_llm_profile",
"description": "Update an LLM profile; only the fields you pass change.",
"inputSchema": {
"type": "object",
"$defs": {
"ProfileAuth": {
"oneOf": [
{
"type": "object",
"required": [
"type"
],
"properties": {
"type": {
"type": "string",
"const": "api_key"
}
}
},
{
"type": "object",
"required": [
"type",
"accessKeyId"
],
"properties": {
"type": {
"type": "string",
"const": "aws_keys"
},
"accessKeyId": {
"type": "string"
}
}
},
{
"type": "object",
"required": [
"type"
],
"properties": {
"type": {
"type": "string",
"const": "bearer_token"
}
}
}
]
},
"ProfileSecrets": {
"type": "object",
"properties": {
"token": {
"type": [
"string",
"null"
]
},
"apiKey": {
"type": [
"string",
"null"
]
},
"headers": {
"type": "object",
"additionalProperties": {
"type": "string"
}
},
"secretAccessKey": {
"type": [
"string",
"null"
]
}
},
"description": "Decrypted contents of `llm_profiles.secrets`; also the plaintext shape the\nwrite routes accept. Absent keys are omitted on the wire (Zod rejects `null`)."
},
"LlmProfileProvider": {
"oneOf": [
{
"enum": [
"openai_completions",
"openai_responses",
"anthropic",
"gemini",
"groq",
"mistral",
"bedrock",
"azure_chat_completions",
"azure_responses",
"azure_anthropic"
],
"type": "string"
},
{
"type": "string",
"const": "custom",
"description": "OpenAI-compatible gateway over Chat Completions."
},
{
"type": "string",
"const": "custom_responses",
"description": "OpenAI-compatible gateway over the Responses API."
}
],
"description": "Wire names shared with the frontend (`LLM_PROFILE_PROVIDERS`)."
},
"ProfileConfigPatch": {
"type": "object",
"properties": {
"auth": {
"anyOf": [
{
"$ref": "#/$defs/ProfileAuth"
},
{
"type": "null"
}
],
"default": null
},
"region": {
"type": [
"string",
"null"
],
"default": null,
"description": "`\"\"` clears it; likewise `resourceId`, `baseUrl` and `apiVersion`."
},
"baseUrl": {
"type": [
"string",
"null"
],
"default": null
},
"apiVersion": {
"type": [
"string",
"null"
],
"default": null
},
"resourceId": {
"type": [
"string",
"null"
],
"default": null
},
"headerNames": {
"type": [
"array",
"null"
],
"items": {
"type": "string"
},
"default": null,
"description": "Custom header names, with their values in `secrets.headers`. `[]` removes every header\nand its stored value."
}
},
"description": "The `config` fields to change. The service replaces `config` whole, so a field a\nschema-driven client sends as `null` would otherwise reset it (and `headerNames` would drop\nthe stored header secrets with it)."
}
},
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"profileId"
],
"properties": {
"name": {
"type": [
"string",
"null"
],
"default": null
},
"config": {
"anyOf": [
{
"$ref": "#/$defs/ProfileConfigPatch"
},
{
"type": "null"
}
]
},
"models": {
"type": [
"array",
"null"
],
"items": {
"type": "string"
},
"default": null,
"description": "Replaces the whole list."
},
"secrets": {
"$ref": "#/$defs/ProfileSecrets",
"default": {},
"description": "Absent keys, or the masks `get_llm_profile` returns, keep the stored secret."
},
"provider": {
"anyOf": [
{
"$ref": "#/$defs/LlmProfileProvider"
},
{
"type": "null"
}
],
"default": null,
"description": "Changing it requires `config`, which then starts empty instead of from the stored one."
},
"profileId": {
"type": "string",
"format": "uuid"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
}
}
}
}
The title clearly communicates the tool’s action and subject. Keep it concise.
Description73/100
Update a signal; only the fields you pass change. Changing `prompt` or
`structuredOutput` adds a definition version.
The description explains the capability, when to use it, and what to expect.
Inputs32/100
5 of 15 inputs have descriptions. Add descriptions to the missing inputs, including where to obtain IDs and how to supply values.
signalId
Not provided
Missing description. Explain what this value means and how to supply it.
name
Not provided
Missing description. Explain what this value means and how to supply it.
prompt
Not provided
Missing description. Explain what this value means and how to supply it.
structuredOutput
Not provided
Missing description. Explain what this value means and how to supply it.
sampleRate
Percent of matching traces to evaluate, 1-95.
Remove validation rules already expressed in this field’s schema; describe its meaning and usage instead.
disabled
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › Trigger › oneOf › 0 › type
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › Trigger › oneOf › 1 › spanNames
Not provided
Missing description. Explain what this value means and how to supply it.
$defs › Trigger › oneOf › 1 › type
Not provided
Missing description. Explain what this value means and how to supply it.
filters
Absent = leave stored; `[]` = run on every firing trace.
Clear enough to supply this argument.
mode
Not provided
Missing description. Explain what this value means and how to supply it.
llmProfileId
Absent = leave stored; set together to re-route the signal.
Clear enough to supply this argument.
model
Not provided
Missing description. Explain what this value means and how to supply it.
clearSampleRate
Stop sampling, so the signal runs on every matching trace. `sampleRate: null` does not.
Clear enough to supply this argument.
projectId
Laminar project to act on. Required when signed in with OAuth: call `list_projects` and
pass the project's id. Ignored for a project API key, which is bound to its project.
Clear enough to supply this argument.
Annotations0/100
Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior.
Not declared
Output schema0/100
No output schema is declared. Add one for structured results so clients can interpret returned fields.
Not declared
Original tool metadata
{
"name": "update_signal",
"description": "Update a signal; only the fields you pass change. Changing `prompt` or\n`structuredOutput` adds a definition version.",
"inputSchema": {
"type": "object",
"$defs": {
"Mode": {
"enum": [
"batch",
"realtime"
],
"type": "string"
},
"Trigger": {
"oneOf": [
{
"type": "object",
"required": [
"type"
],
"properties": {
"type": {
"type": "string",
"const": "rootSpanFinished"
}
},
"description": "The trace's root span finished. Right for most traces."
},
{
"type": "object",
"required": [
"type",
"spanNames"
],
"properties": {
"type": {
"type": "string",
"const": "spanName"
},
"spanNames": {
"type": "array",
"items": {
"type": "string"
}
}
},
"description": "A span with any of these names finished. For distributed traces where no\nsingle span is observably the root."
}
],
"description": "WHEN a signal is evaluated, answerable from a single span batch. A closed set\nrather than a `{column, operator, value}` list: these are the only shapes the\nevaluator implements, so anything else was stored and then never fired."
}
},
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"signalId"
],
"properties": {
"mode": {
"anyOf": [
{
"$ref": "#/$defs/Mode"
},
{
"type": "null"
}
],
"default": null
},
"name": {
"type": [
"string",
"null"
],
"default": null
},
"model": {
"type": [
"string",
"null"
],
"default": null
},
"prompt": {
"type": [
"string",
"null"
],
"default": null
},
"filters": {
"type": [
"array",
"null"
],
"items": true,
"default": null,
"description": "Absent = leave stored; `[]` = run on every firing trace."
},
"trigger": {
"anyOf": [
{
"$ref": "#/$defs/Trigger"
},
{
"type": "null"
}
],
"default": null,
"description": "Absent or `null` → leave stored."
},
"disabled": {
"type": [
"boolean",
"null"
],
"default": null
},
"signalId": {
"type": "string",
"format": "uuid"
},
"projectId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Laminar project to act on. Required when signed in with OAuth: call `list_projects` and\npass the project's id. Ignored for a project API key, which is bound to its project."
},
"sampleRate": {
"type": [
"integer",
"null"
],
"format": "int64",
"default": null,
"description": "Percent of matching traces to evaluate, 1-95."
},
"llmProfileId": {
"type": [
"string",
"null"
],
"format": "uuid",
"default": null,
"description": "Absent = leave stored; set together to re-route the signal."
},
"clearSampleRate": {
"type": "boolean",
"default": false,
"description": "Stop sampling, so the signal runs on every matching trace. `sampleRate: null` does not."
},
"structuredOutput": {
"default": null
}
}
}
}
Loading score analytics…
Put your MCP to the test
A few tasks worth trying with this server, chosen from its reviewed tools. These have not been run.
01List projectsList the Laminar projects this connection can act on, with their workspaces.
02Query laminar sqlExecute a read-only SQL query (full ClickHouse SELECT syntax) against this project's Laminar trace data and get rows back as a JSON array.
03Ask agentAsk Laminar Agent a question about this project's traces in plain language and get a
04List signalsList the project's signals (LLM analyzers that run on matching traces) with their prompt,
05Get trace contextGet an LLM-optimized summary of a trace.
Want to run these tasks against your MCP server with us?
laminar setup, tools, quality score, publishing readiness, and alternatives.
How do I add laminar to Claude, ChatGPT, or Cursor?
Add laminar to Claude, ChatGPT, or Cursor by connecting its remote MCP URL in each client's MCP settings. In Claude, add a custom connector. In ChatGPT, add an MCP app. In Cursor, add an MCP server under Settings. The MCP URL is https://api.lmnr.ai/v1/mcp.
Is laminar MCP ready for ChatGPT and Claude?
As of November 19, 2026, laminar scores 78/100 on Manufact's MCP publishing checklist. The scan does not yet meet ChatGPT publishing checks and does not yet meet Claude publishing checks. Passing these checks does not guarantee marketplace approval.
How clear is laminar's MCP tool documentation?
As of November 19, 2026, laminar scores 54/100 in Manufact's review of tool titles, descriptions, input and output schemas, and behavior annotations. The review covers 16 of 16 discovered tools. The quality score measures documentation clarity; runtime behavior and marketplace readiness require separate checks.
What are the best alternatives to laminar?
As of November 20, 2026, alternatives to laminar in developer-tools on the Manufact MCP leaderboard include Speko, Manufact, AgentMail, Context, and AgentPhone.
How can laminar improve its MCP documentation?
create_llm_profile is the lowest-scoring reviewed tool at 34/100. Declare readOnlyHint, destructiveHint, idempotentHint, openWorldHint based on the tool’s actual behavior. Open its review for feedback and a copyable fix prompt, then rerun the documentation review after updating your tool metadata.
Not your MCP?
Review any public or OAuth MCP server for protocol compatibility, tool documentation, and ChatGPT and Claude publishing readiness. No account required.