polarion-mcp

What it does

Everything Polarion, AI-native

One MCP server that turns the entire Polarion REST API into tool calls your AI assistant can use directly.

🔌

284 REST Operations

Every Polarion REST endpoint exposed as a typed tool — work items, test runs, documents, plans, users, and more. Full coverage, auto-generated from the OpenAPI spec.

auto-generated

🤖

Every Major AI Client

Claude Code, Claude Desktop, VS Code GitHub Copilot, ChatGPT Custom GPT, Claude.ai web — one server, five clients, three transports. Pick your stack.

stdio · http · streamable

🐳

Docker-Ready

Multi-transport container image with configurable entrypoints. Run locally or deploy on your own infrastructure. No cloud dependency.

self-hosted

📚

SDK Documentation Access

Optional Polarion SDK docs loaded as tool context. Your AI understands not just the REST API, but the Polarion programming model.

in-context docs

🔒

Secure by Design

Your Polarion Personal Access Token stays yours — in a gitignored .env for stdio, sent per-request for HTTP, or entered once through an OAuth login: the server stores nothing and the client only ever holds it encrypted, unreadable to anyone but the server.

pat · oauth · tls

⚡

Cached Project Config

Project-level configuration is cached on first use. Your AI stays fast even when exploring a complex Polarion instance with dozens of projects.

performance

Quick Start

Get connected

Choose your AI client. Copy the config. You're done.

claude mcp add polarion-mcp \
  -e API_BASE_URL=https://your-polarion/polarion/rest/v1 \
  -e BEARER_TOKEN=your-personal-access-token \
  -- node /path/to/polarion-mcp/build/index.js

Local stdio: Claude Code starts the server itself, nothing listens on the network. API_BASE_URL always ends in /polarion/rest/v1. Create the token in Polarion under My Account › Personal Access Tokens.

// .vscode/mcp.json (VS Code asks for the token and stores it securely)
{
  "inputs": [
    { "type": "promptString", "id": "polarion-pat", "description": "Polarion PAT", "password": true }
  ],
  "servers": {
    "polarion-mcp": {
      "type": "stdio",
      "command": "node",
      "args": ["/path/to/polarion-mcp/build/index.js"],
      "env": {
        "API_BASE_URL": "https://your-polarion/polarion/rest/v1",
        "BEARER_TOKEN": "${input:polarion-pat}"
      }
    }
  }
}

// claude_desktop_config.json, then restart Claude Desktop
{
  "mcpServers": {
    "polarion-mcp": {
      "command": "node",
      "args": ["/path/to/polarion-mcp/build/index.js"],
      "env": {
        "API_BASE_URL": "https://your-polarion/polarion/rest/v1",
        "BEARER_TOKEN": "your-personal-access-token"
      }
    }
  }
}

No local Node? Use "command": "docker" with "args": ["run", "--rm", "-i", "-e", "API_BASE_URL", "-e", "BEARER_TOKEN", "ghcr.io/phillipboesger/polarion-mcp:latest", "node", "build/index.js"].

docker run -d --name polarion-mcp \
  -e API_BASE_URL=https://your-polarion/polarion/rest/v1 \
  -e MCP_PUBLIC_URL=https://mcp.yourdomain.com \
  -e MCP_TOKEN_SECRET=$(openssl rand -base64 48) \
  -e MCP_ALLOWED_HOSTS=mcp.yourdomain.com \
  -p 127.0.0.1:3000:3000 \
  ghcr.io/phillipboesger/polarion-mcp:latest

# connector URL for Claude.ai and ChatGPT
URL:  https://mcp.yourdomain.com/mcp

For clients that cannot send a custom header. Put the server behind TLS on a host reachable from the internet (Claude.ai and ChatGPT connect from their cloud, not from your network). Claude.ai: Settings › Connectors › Add custom connector, paste the URL, leave client ID and secret empty. ChatGPT: turn on Developer mode under Settings › Security and login, add a connector with the URL and OAuth. On first use each user pastes their own Polarion PAT on a login page.

The server stores no token: the client receives it encrypted under MCP_TOKEN_SECRET, which only the server can read. Access renews silently every hour and each renewal re-checks the PAT with Polarion, so a login lasts until the PAT is revoked or expires, and survives server restarts. Keep MCP_TOKEN_SECRET fixed; changing it logs everyone out. Reverse-proxy paths: docs/deployment.md.

# the OAuth server from the previous tab, plus a client for the GPT
-e GPT_CLIENT_ID=polarion-gpt \
-e GPT_CLIENT_SECRET=$(openssl rand -hex 32)

# GPT editor: Actions › Import from URL
https://mcp.yourdomain.com/openapi-gpt.json

# Authentication › OAuth
Client ID:          polarion-gpt
Client Secret:      the GPT_CLIENT_SECRET value
Authorization URL:  https://mcp.yourdomain.com/gpt/authorize
Token URL:          https://mcp.yourdomain.com/gpt/token
Scope:              polarion

Each GPT user signs in once with their own Polarion PAT and then works under their own Polarion account; the login lasts as long as that PAT is valid. Either token exchange method (POST or Basic) works. Only ChatGPT's own /aip/<gpt-id>/oauth/callback addresses are accepted, so nothing needs allowlisting.

docker run -d --name polarion-mcp \
  -e API_BASE_URL=https://your-polarion/polarion/rest/v1 \
  -p 127.0.0.1:3000:3000 \
  ghcr.io/phillipboesger/polarion-mcp:latest

# Claude Code, behind your TLS proxy
claude mcp add --transport http polarion https://mcp.yourdomain.com/mcp \
  --header "Authorization: Bearer <your-polarion-pat>"

One shared server that holds no credentials; each client sends its own Polarion PAT as a header, and Polarion authorizes every call under that user. For Claude Code, VS Code Copilot ("type": "http" with "headers"), curl and scripts. Claude.ai and ChatGPT connectors cannot set a header: use the OAuth tab instead.

docker run -d \
  -e API_BASE_URL=https://your-polarion/polarion/rest/v1 \
  -e BEARER_TOKEN=service-account-pat \
  -e HTTP_API_KEY=$(openssl rand -hex 32) \
  -p 127.0.0.1:3000:3000 \
  ghcr.io/phillipboesger/polarion-mcp:latest \
  node build/http-server.js

A plain REST wrapper for Custom GPT Actions with one shared key: every GPT user acts as the same Polarion account. In the GPT editor import https://your-host/openapi-gpt.json (a 30-tool subset that fits ChatGPT's limit) and set authentication to API key, type Bearer, with the HTTP_API_KEY value. For per-user logins use the Custom GPT (OAuth) tab.

Your AI, inside Polarion

Everything Polarion, AI-native

Works with your tools

Up and running in 3 steps

Get connected

Built by Phillip Bösger

Ready to connect your AI to Polarion?