@iterium/mcp-server

Connects Claude Desktop, Cursor, or any stdio-based MCP client to your deployed Iterium instance.

This package is a thin bridge, not a reimplementation: it discovers the tools exposed by your Iterium deployment's own MCP endpoint (/api/mcp, Streamable HTTP) and forwards every request to it verbatim. No API keys live in this package — your Jira/GitHub/Confluence/Figma credentials stay server-side, exactly where Iterium's Integrations tab already has them configured.

Requirements

An Iterium instance deployed somewhere reachable over HTTPS (e.g. on Vercel), with its MCP endpoint at <your-deployment>/api/mcp.

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "iterium": {
      "command": "npx",
      "args": ["-y", "@iterium/mcp-server"],
      "env": {
        "ITERIUM_MCP_URL": "https://your-team.vercel.app/api/mcp"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json (same shape):

{
  "mcpServers": {
    "iterium": {
      "command": "npx",
      "args": ["-y", "@iterium/mcp-server"],
      "env": { "ITERIUM_MCP_URL": "https://your-team.vercel.app/api/mcp" }
    }
  }
}

Environment variables

| Variable | Required | Description | | --- | --- | --- | | ITERIUM_MCP_URL | Yes | Your deployment's MCP endpoint, e.g. https://your-team.vercel.app/api/mcp. | | ITERIUM_MCP_TOKEN | No | Sent as Authorization: Bearer <token> if your deployment protects /api/mcp. |

What's exposed

Whatever your deployed Iterium instance registers in src/lib/mcp-server.ts — today that's 8 read-oriented tools (Jira, Confluence, GitHub, Figma, Sprint Intelligence). Add a tool there and it appears here automatically on the next connection, no republish needed.

Local development

pnpm --filter @iterium/mcp-server build
ITERIUM_MCP_URL=http://localhost:3000/api/mcp node mcp/dist/stdio.js