Polarion MCP Server

A TypeScript MCP (Model Context Protocol) server that turns the Polarion ALM REST API into a tool-based interface for AI assistants — Claude Code, Claude Desktop, VS Code (GitHub Copilot), and any other MCP-aware client. It also ships an optional HTTP wrapper so HTTP-only clients such as ChatGPT Custom GPTs can call the same tools.

Note: Independent, community open-source project. Not affiliated with or endorsed by Siemens / Polarion. "Polarion" is a trademark of Siemens. Polarion SDK/help PDFs are not bundled — see sdk/README.md.

What you get

Three transports, one shared tool set (~210 Polarion REST operations):

• stdio MCP — for local clients that launch the server as a subprocess (Claude Code, Claude Desktop, VS Code).

• Streamable HTTP MCP (/mcp) — a real remote MCP endpoint for clients that connect by URL, such as Claude.ai custom connectors.

• REST HTTP wrapper — a plain REST surface for ChatGPT Custom GPT Actions.

Plus Polarion-specific extras: cached project configuration, SDK documentation access, and guided prompts for creating/searching/updating work items.

Related MCP server: MCP-Typescribe

Contents

• Prerequisites

• Get a Polarion token

• Build it once

• Install in your AI client

  • Claude Code (CLI)

  • Claude Desktop

  • VS Code (GitHub Copilot)

  • ChatGPT (Custom GPT)

  • Claude.ai (web)

• Run with Docker

• Configuration reference

• Security notes

• Documentation

Prerequisites

• Node.js 20+ and npm

• Access to a Polarion instance with the REST API enabled

• A Polarion Personal Access Token (see below)

Get a Polarion token

Set BEARER_TOKEN to a Polarion Personal Access Token:

1. Sign in to your Polarion instance.

2. Open your user profile / personal settings.

3. Find the Personal Access Token section.

4. Create a token with the minimum scope your workflow needs.

5. Copy it once and store it in your MCP client config (or .env).

If your instance does not show Personal Access Tokens, the feature may be disabled — contact your Polarion administrator.

Build it once

All local (stdio) integrations use the compiled build/index.js. Clone and build first:

git clone https://github.com/phillipboesger/polarion-mcp.git
cd polarion-mcp
npm install      # installs deps and builds (via the prepare script)
npm run build    # (re)build explicitly if needed

Note the absolute path to build/index.js — you will paste it into each client config below. Get it with:

echo "$(pwd)/build/index.js"

Two environment variables are required for every client:

• API_BASE_URL — e.g. https://your-polarion-server/polarion/rest/v1

• BEARER_TOKEN — your Polarion Personal Access Token

For internal servers with a self-signed certificate, also set NODE_TLS_REJECT_UNAUTHORIZED=0.

Install in your AI client

Claude Code (CLI)

Use the claude mcp add command (replace the path and values):

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

Then verify:

claude mcp list

To scope the server to one project only, run the command from that project directory (Claude Code stores it in the project's local config), or add --scope user to make it available everywhere.

Claude Desktop

Edit the Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "polarion": {
      "command": "node",
      "args": ["/absolute/path/to/polarion-mcp/build/index.js"],
      "env": {
        "API_BASE_URL": "https://your-polarion-server/polarion/rest/v1",
        "BEARER_TOKEN": "your-polarion-personal-access-token"
      }
    }
  }
}

Restart Claude Desktop. The Polarion tools appear under the 🔌 (connectors) icon.

VS Code (GitHub Copilot)

VS Code runs MCP servers in Copilot agent mode. Create .vscode/mcp.json in your workspace (or add the same block to your user settings.json under "mcp"):

{
  "servers": {
    "polarion": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/polarion-mcp/build/index.js"],
      "env": {
        "API_BASE_URL": "https://your-polarion-server/polarion/rest/v1",
        "BEARER_TOKEN": "your-polarion-personal-access-token"
      }
    }
  }
}

Open the Chat view, switch to Agent mode, and start the polarion server from the tools picker. You can run several entries (e.g. polarion-sim, polarion-dev) with different API_BASE_URL/BEARER_TOKEN values to target multiple environments.

ChatGPT (Custom GPT)

ChatGPT cannot launch a local MCP server, so use the HTTP wrapper and wire it up as a Custom GPT Action.

1. Deploy the HTTP server (see Run with Docker or docs/deployment.md) with API_BASE_URL, BEARER_TOKEN, and a generated HTTP_API_KEY:

   bash scripts/generate-api-key.sh   # prints a secure HTTP_API_KEY
   npm run start:http

2. In ChatGPT → Create a GPT → Configure → Actions → Import from URL, point at your deployment's /openapi-gpt.json (a Custom-GPT-friendly spec limited to 30 operations).

3. Set Authentication → API Key → Bearer, and paste the HTTP_API_KEY value.

4. Save. The GPT can now call the Polarion tools over HTTPS.

The full /openapi.json (all ~210 operations) is also available, but exceeds the 30-action Custom GPT limit — use /openapi-gpt.json there.

Claude.ai (web)

Claude.ai (Pro/Max/Team/Enterprise) supports custom connectors — remote MCP servers reached over HTTPS. This repo ships a real Streamable HTTP MCP transport for exactly this.

1. Generate a token and start the MCP HTTP server (host it behind HTTPS — see Run with Docker and docs/deployment.md):

   export MCP_HTTP_TOKEN="$(openssl rand -hex 32)"
   export API_BASE_URL="https://your-polarion-server/polarion/rest/v1"
   export BEARER_TOKEN="your-polarion-personal-access-token"
   export MCP_ALLOWED_HOSTS="mcp.example.com"   # optional: DNS-rebinding protection
   npm run start:mcp-http                        # serves POST/GET/DELETE /mcp

2. Expose it publicly over HTTPS, e.g. https://mcp.example.com/mcp.

3. In Claude.ai → Settings → Connectors → Add custom connector, enter the /mcp URL.

4. Auth: the endpoint always requires Authorization: Bearer <MCP_HTTP_TOKEN> (the server refuses to start without MCP_HTTP_TOKEN). If the Claude.ai connector dialog cannot send a static bearer header for your plan, front the server with an OAuth-capable proxy or restrict it at the network layer.

Prefer no hosting? Claude Desktop (above) uses the same Claude account over local stdio and needs no public endpoint.

Run with Docker

Build the image:

docker build -t polarion-mcp .

HTTP mode (default command — for ChatGPT / hosted use):

docker run --rm -p 3000:3000 \
  -e API_BASE_URL="https://your-polarion-server/polarion/rest/v1" \
  -e BEARER_TOKEN="your-polarion-personal-access-token" \
  -e HTTP_API_KEY="your-generated-http-key" \
  polarion-mcp

Streamable HTTP MCP mode — for Claude.ai (override the command):

docker run --rm -p 3000:3000 \
  -e API_BASE_URL="https://your-polarion-server/polarion/rest/v1" \
  -e BEARER_TOKEN="your-polarion-personal-access-token" \
  -e MCP_HTTP_TOKEN="your-generated-mcp-token" \
  polarion-mcp node build/mcp-http-server.js

stdio MCP mode (override the command):

docker run --rm -i \
  -e API_BASE_URL="https://your-polarion-server/polarion/rest/v1" \
  -e BEARER_TOKEN="your-polarion-personal-access-token" \
  polarion-mcp node build/index.js

Endpoints — Streamable HTTP MCP: POST/GET/DELETE /mcp (+ GET /health). REST wrapper: GET /health, GET /api/tools, POST /api/tools/:toolName, GET /openapi.json, GET /openapi-gpt.json.

Configuration reference

Copy .env.example to .env for local development. See docs/configuration.md for advanced per-scheme auth overrides.

Security notes

• Never commit tokens. .env is gitignored.

• Keep MCP configs in user-level settings or ignored paths.

• Use NODE_TLS_REJECT_UNAUTHORIZED=0 only for trusted internal servers.

• Use a distinct HTTP_API_KEY for the HTTP wrapper — it is not the Polarion bearer token.

Documentation

See docs/README.md for architecture, features, workflows, configuration, deployment, security, and troubleshooting guides. A worked end-to-end example lives in docs/example-task.md.

Contributing

Contributions welcome — see CONTRIBUTING.md.

License

MIT