MCP Overview

How the Fenix MCP bridge connects your AI agents to your workspace.

fenix-mcp is a Python server that implements the Model Context Protocol (MCP). It runs locally as a bridge between your AI agents and your Fenix workspace, exposing 12 tools and 68+ actions over stdio or HTTP transport.

Why MCP?

Most project management tools are passive. You open them in a browser, copy-paste context into your AI chat, and lose track of decisions between sessions.

Fenix MCP makes your workspace an active participant:

Your AI reads real sprint data instead of guessing
Decisions are saved as searchable memories between sessions
Tasks can be created, updated, and moved from within your agent
Documentation is written and published without leaving the terminal

Agent (Claude / Cursor / Copilot / …)
         │
         │ MCP stdio / HTTP
         ▼
   fenix-mcp server (local)
         │
         │ HTTPS REST
         ▼
   Fenix Cloud API
         │
         ├── Work Items
         ├── Memories (pgvector embeddings)
         ├── Docs / Wiki
         ├── Sprints & Boards
         ├── Skills & Rules
         └── API Catalog

Supported agents

Agent	Transport
Claude Code	stdio
Cursor	stdio
Windsurf	stdio
GitHub Copilot	stdio
OpenAI Codex	stdio
Kiro	stdio
Gemini CLI	stdio
Aider	stdio

The 12 MCP tools

Tool	Description	Actions
`health`	Backend health check	1
`initialize`	Init environment, load user context	1
`team`	Switch teams, list teams	3
`work_items`	Create, list, update, move status	11
`docs`	Create, read, update, publish	14
`productivity`	Personal TODOs	11
`intelligence`	Save and search semantic memories	4
`sprints`	List, active sprint, items	4
`boards`	View boards and columns	4
`skills`	Install and export skills	11
`api_catalog`	Semantic search over API docs	3
`user_config`	Personal config documents	6

Authentication

Personal Access Token — set FENIX_API_KEY=fxpat_... in your environment.

OAuth 2.1 — handled automatically when using fenix-plugin with Claude Code.

Transport modes

fenix-mcp                           # stdio (default)
fenix-mcp --http --port 8421        # HTTP server
fenix-mcp --http --port 8421 --stdio # both