FlowMCP

What is FlowMCP?

FlowMCP is a framework that adapts existing web APIs (e.g., REST) into a standardized Model Context Protocol (MCP) interface, enabling AI systems to consume those APIs in a structured, testable, and semantically consistent way. It abstracts complex APIs into AI-friendly schema definitions and provides built-in tools for testing, validation, and pre/post processing.

How to use FlowMCP?

Install the package, create a schema (.mjs file) following the documented structure, then activate it using FlowMCP.activateServerTools({ server, schema, serverParams }) in a Node.js script that creates an MCP server with StdioServerTransport. Environment variables (e.g., API keys) are passed via serverParams. Pre-built schemas are available in the community repository.

Core features of FlowMCP

• Schema-based integration with AI-optimized parameters
• Pre/post processing modifiers for query/results
• Built-in automated testing for routes
• Text-based human-readable output with error messages
• Zod validation for parameter inputs
• Support for GET and POST requests with dynamic routes

Use cases of FlowMCP

• Expose a REST API as MCP tools for AI assistants
• Create AI-friendly wrappers for data providers like CoinGecko or Etherscan
• Build testable, schema-driven integrations for blockchain or DeFi data
• Enable structured, validated queries from AI systems to external APIs

FAQ from FlowMCP

What dependencies does FlowMCP require?

FlowMCP depends on the @modelcontextprotocol/sdk, Node.js, and the Zod validation library. It runs in any Node.js environment and connects via the standard I/O transport.

How are authentication credentials handled?

API keys and other secrets are passed as environment variables or via the serverParams object during schema activation, using {{ENV_VAR}} placeholders in the schema definition.

What transport or protocol does FlowMCP use?

FlowMCP uses the Model Context Protocol with the StdioServerTransport; all communication happens over standard input/output.

Are there any limits on schema definitions?

Yes. The namespace key must be unique, alphanumeric only, and at most 24 characters. Route definitions require either GET or POST methods and must include a route path with optional dynamic inserts.

Where do schemas and data live?

Schemas are local .mjs files in the project. Data is fetched from the external APIs defined in the root URL. No data is stored by FlowMCP itself.