AI Workflow MCP

Sign in

Docs

Overview

AI Workflow MCP is a Cloudflare-native MCP server for AI coding facts.

AI Workflow MCP gives coding agents a typed source for facts they otherwise waste time searching for: model pricing, context windows, model capabilities, AI coding tool metadata, official docs, Skills, public MCP servers, thinking frameworks, and freshness state.

The production endpoint is:

https://mcp.aiworkflowmcp.com/mcp

Agent Contract

AI Workflow MCP is English-first at the protocol layer. Tool names, JSON fields, evidence metadata, identifiers, and Markdown payloads stay English even when the marketing site is localized.

Agents may translate non-English user requests into concise English before calling search-like tools. UI copy can localize navigation, but service contracts stay stable for Claude Code, Codex, Cursor, Cline, Windsurf, OpenCode, Gemini CLI, and other clients.

Core Tools

  • get_ai_model: model pricing, context, modalities, limits, provenance, and evidence.
  • get_coding_tool: install, command surface, supported models, environment variables, and source evidence.
  • search_official_docs: official docs with optional full Markdown loaded from R2.
  • search_skills: official and major-vendor agent Skills.
  • search_cli: CLI command-surface facts.
  • search_mcp: official and major-vendor MCP servers.
  • search_frameworks, route_framework, get_framework: operational thinking frameworks.

Response Shape

Every successful MCP tool response returns:

  • content: a short text summary for agent chat surfaces.
  • structuredContent: typed JSON for deterministic downstream use.
  • _meta: source, freshness, evidence, and next-step metadata where available.

Every public fact should trace back to an official source URL, content hash, and evidence key. Commercial retrieval paths should use fresh_only=true when stale facts must be rejected.

Production Path

Use this order when wiring a real user or agent:

  1. Create an account at https://aiworkflowmcp.com/en/sign-in/.
  2. Create an API key at https://aiworkflowmcp.com/en/account/api-keys/.
  3. Add https://mcp.aiworkflowmcp.com/mcp to the agent client.
  4. Send Authorization: Bearer ${AIWORKFLOWMCP_API_KEY} for authenticated MCP and REST calls.
  5. Check https://mcp.aiworkflowmcp.com/health before high-volume or freshness-sensitive workflows.

Public calls work without a key for basic exploration. Authenticated calls add account-level usage history and higher per-minute quota routing.