MCP research notes¶
What is it?¶
A: It’s effectively just OpenAPI for LLMs._
A: It is sometimes described as “the USB-C port for AI”, providing a uniform way to connect LLMs to resources they can use.
Specification¶
The MCP specification.
Introduction¶
MCP: an in-depth introduction https://www.speakeasy.com/mcp/mcp-tutorial https://news.ycombinator.com/item?id=43972334
Evolution¶
The spec for the MCP transport is changing from SSE to Streamable HTTP.
Gallery¶
Articles¶
What we learned when converting complex OpenAPI specs to MCP servers https://www.stainless.com/blog/what-we-learned-converting-complex-openapi-specs-to-mcp-servers
Package registries¶
Clients¶
Servers¶
Frameworks¶
Databases¶
PostgreSQL - Read-only database access with schema inspection, see @modelcontextprotocol/server-postgres.
DBHub - Universal database MCP server connecting to MySQL, PostgreSQL, SQLite, DuckDB and etc.
JDBC - Connect to any JDBC-compatible database and query, insert, update, delete, and more. Supports MySQL, PostgreSQL, Oracle, SQL Server, sqllite and more.
Copilots¶
Docs and Knowledge¶
Mastra MCP Documentation Server
https://mintlify.com/blog/generate-mcp-servers-for-your-docs
Others¶
https://github.com/lastmile-ai/mcp-agent
mcp-agent is a simple, composable framework to build agents using Model Context Protocol. This one also looks interesting, including something about marimo.
https://openai.github.io/openai-agents-python/mcp/ (discuss)
https://github.com/microsoft/semanticworkbench/tree/main/mcp-servers