FlowMCP is a framework designed to adapt and expose existing web APIs (e.g., REST interfaces) through a standardized Model Context Protocol (MCP) interface. It allows APIs to be consumed by AI systems in a structured, testable, and semantically consistent way.
Documentation →
- Hackathon Builder — Copy the Hackathon Kit skill snippet into your AI assistant's context and scaffold an MCP server in <60s. Nothing to install.
- AI Engineer — Architecture in 30 min — Concepts + Specification + Reference.
- Schema Maintainer — Strict v4.3.0 schema format with shared lists + validation rules.
- Decision Maker — Use cases, comparison, trust, roadmap — in 5 min.
- flowmcp-cli: Command-line interface for developing, validating, and managing FlowMCP API schemas
- flowmcp-core: Framework for adapting REST APIs into MCP-compatible tools with schema-driven validation
- flowmcp-schemas: — curated API schemas across — unique datasources
- flowmcp-spec: FlowMCP v4.3.0 specification — 24 spec documents, reference examples, LLM-consumable llms.txt
- flowmcp-servers: Local (stdio) and remote (HTTP/SSE) servers for deploying FlowMCP schemas
- flowmcp-grading: Schema-quality grading system — assesses single-schema and selection-level quality across phases, dimensions, and tiers (autonomous / group-bound)
- x402-core: Multi-chain ERC20 payment layer using EIP-3009 signed authorizations with CAIP-2 network routing
- x402-mcp-middleware: Express middleware for payment-gated MCP servers with automatic routing and JSON-RPC 402 compliance
- x402-flowmcp-org: Payment-gated MCP server with X402 on-chain payments, Avalanche data, and A2A protocol support
- mcp-agent-server: MCP server with AI agent-powered tools built on FlowMCP schemas
- AgentProbe: Multi-protocol agent endpoint validator — own GitHub organization
- geo-gtfs-toolkit: Convert GTFS Schedule feeds (CSV in ZIP) to queryable SQLite with quality seal, capability detection, and reusable default queries
- geo-geojson-toolkit: Load GeoJSON FeatureCollections (RFC 7946) from a URL into memory and serve spatially queryable, auto-injected spatial tools
- geo-csv-tsv-toolkit: Load geo CSV/TSV from a URL into memory with a mandatory parse config (no silent defaults) and auto-injected spatial tools
- geo-overpass-toolkit: Query OpenStreetMap via the Overpass API as a live-query source — curated multi-key categories, pre-built combined selections, and reusable spatial methods (nearPoint, inBoundingBox, discoverCategories, runOverpassQL)
- time-csv-toolkit: Load event CSVs from a URL into memory with a mandatory parse config (no silent defaults) and run deterministic, structured date methods (addInterval, eventsInRange, eventsInCalendarWeek, timeResolve) with additive ISO-8601 normalization
- geo-dzt-toolkit: Query the DZT Knowledge Graph (Open Data Germany) as a live-query geo source — SPARQL bbox-FILTER + client-side Haversine (no GeoSPARQL), RFC 7946 lon-first FeatureCollections, plus name/SPARQL search
- geo-zhv-toolkit: Convert the German central stop registry (zHV / Zentrales Haltestellenverzeichnis, DELFI) XML into a sealed SQLite stop directory (DHID + coordinates + AGS) with a heuristic EVA↔DHID bridge
License: MIT