How Alpha Vantage automated MCP hosting and QA with Manufact

Background

Alpha Vantage (YC S18) makes institutional-grade financial data available to anyone who wants to build on it: equities, options, fundamentals, economic indicators, news and sentiment, and more. Its users range from solo developers and retail investors to fintech companies and large enterprises. The company has always competed on developer experience. Finding the right dataset should be easy, and so should integrating it.

MCP gave the team a way to push that further. Before, a developer had to read the API docs, write integration code, parse responses, and build logic on top. With the Alpha Vantage MCP server, an agent can see which financial tools exist, pick the right one, pull the data, and reason over it in the same conversation.

"Instead of asking, 'Which API endpoint do I need?' users could simply ask a financial question and let the agent handle the underlying data orchestration. The vision was to make Alpha Vantage a financial data and intelligence layer for the agentic ecosystem."

Alpha Vantage (YC S18)

The original plan: build it themselves

Alpha Vantage set out to build their MCP the way most API companies would: by standing up its own hosting stack on AWS, Azure, or GCP.

A cloud provider gives you servers. What it doesn't tell you is whether your MCP works in Claude, ChatGPT, Meta Muse, and whichever client launches next month. Alpha Vantage needed one server to work plug-and-play in all of them. Each client handles tools a little differently, so every one meant manual testing, prompt evaluation, and hand-built test suites the team would have to rewrite whenever a client changed. That QA load would have landed on the same engineers who ship the data product, and it would keep growing with every new client.

"The hardest part was to make sure our MCP server could be readily plug and play across a large (and rapidly growing) number of LLM and agentic AI services, such as Claude, ChatGPT, Gemini, and Microsoft Copilot. It involves a lot of manual testing and prompt evaluation."

Alpha Vantage (YC S18)

Why Manufact

Alpha Vantage skipped writing those suites. The team found a fellow YC company already working on their problem: Manufact covers the MCP workflow from design through hosting and evaluation, including cross-client testing that runs one set of checks against ChatGPT, Claude, and other clients.

"Manufact is solving the exact pain points along the MCP design, hosting, and evaluation workflow. We gave it a try and never looked back."

Alpha Vantage (YC S18)

Onboarding took less effort than the team expected. Manufact's engineers worked with Alpha Vantage directly on migrating the server over.

The Result

Alpha Vantage now runs its MCP on Manufact end to end: hosting, feature deployment, cross-client testing, monitoring, and analytics. One platform replaces what would have been a custom cloud stack plus an in-house QA process for every AI client.

With hosting, deployment, and QA automated, the team gets engineering and product cycles back. They spend that time on the financial data product.

"Being able to automate away the key components of MCP hosting, deployment, and QA has saved us a lot of engineering and product cycle time, which has in turn enabled us to focus on the actual product development to drive customer delight."

Alpha Vantage (YC S18)

The Takeaway?

For a data company, an MCP server is a distribution channel in the AI clients its customers already use. Each new client adds another surface to test, and that list grows every quarter. Alpha Vantage chose to keep its engineers focused on market data and let Manufact handle the cross-client work.

Thinking of shipping your own MCP server? Deploy your first server in under 60 seconds at manufact.com, or book a demo and we'll help you map the fastest path to production agent traffic!