Background
Risk, threat and fraud investigations often begin with incomplete information: a company domain, a profile or an identity that does not quite add up. Sixtyfour builds research infrastructure to help teams investigate people and companies, connect evidence across sources and examine what the findings support.
They wanted customers to be able to access those research capabilities through AI assistants and agent workflows. Sixtyfour had already built an MCP server when they came to Manufact for help with the maintenance around that server: publishing checks, submitting to ChatGPT and Claude, client compatibility, hosting, and deployment.
The problem
Building the MCP server was the first step. Sixtyfour also needed to keep up with all of the maintenance: meeting ChatGPT and Claude directory requirements, working through authentication and differences in client behavior, all while developing its own product. Manufact helped take on that work.
The connection needed to work for customers and directory reviewers, as well as in Sixtyfour’s own tests.
“For us, exposing capabilities through MCP meant working through deployment, client compatibility and ecosystem requirements alongside our own research tooling.”
The Sixtyfour team
Why Manufact
A clear set of issues to address
Manufact’s publishing checks gave Sixtyfour’s engineers specific issues to work through before submitting to ChatGPT and Claude. The team resolved the checklist findings, turning the preparation work into a concrete set of changes.
Manufact paired those checks with hands-on submission support. We helped Sixtyfour navigate directory requirements and work through authentication, reviewer access and client behavior.
Hosting and visibility into MCP calls
Alongside publishing support, Manufact hosts Sixtyfour’s MCP server and supports previews and compliance checks on each deployment. Sixtyfour has connected runtime logs to its monitoring, with tracing that follows MCP calls through to its backend.
That gives its engineers visibility across the path between an AI client and the underlying research tools when they investigate an issue. Manufact supports the MCP observability layer, while Sixtyfour develops the research agents, APIs and evidence those tools return.
The result
Sixtyfour’s MCP server runs on Manufact, with deployment support and tracing that follows calls from AI clients through to Sixtyfour’s backend. Manufact also helped the team address publishing-check findings and navigate the submission process for ChatGPT and Claude directories.
In one investigation conducted through its MCP server, Sixtyfour examined an investment vehicle and identified gaps in ownership and funding that required further documentation. It also found no evidence linking a named individual to the vehicle and ruled out a court case involving someone with the same name. The findings distinguished open questions requiring further investigation from an unrelated match. Sixtyfour provided the research and findings, with Manufact providing the hosting and gateway.
“That practical support mattered because MCP is an access point into Sixtyfour’s intelligence infrastructure. It needs to work within the environments our customers and their agents use.”
The Sixtyfour team
The takeaway
Sixtyfour’s longer-term vision includes investigations involving people, businesses and machines, including fraud carried out through automated actors. As that work develops, MCP offers another access point into its research infrastructure.
Manufact’s role is to help prepare that access point for review, support client compatibility, and provide hosting, deployment support and visibility into runtime behavior. For teams building products alongside an MCP server, Sixtyfour’s experience shows where Manufact can contribute.
Explore Sixtyfour’s research infrastructure for risk, threat and fraud investigations.
Building an MCP server alongside your core product? Talk to Manufact about hosting, deployment support and getting ready for review, or start with readiness checks against your existing endpoint.











