Feature request: MCP (Model Context Protocol) server for elmclient

[filhocf]

Summary

Add a built-in MCP (Model Context Protocol) server that exposes elmclient functionality as tools for AI assistants (Claude, GPT, Copilot, etc.).

Motivation

MCP is becoming the standard protocol for connecting AI assistants to external tools and data sources. An MCP server for elmclient would allow AI assistants to directly interact with ELM (EWM, DOORS Next, ETM) — listing projects, querying work items, reading requirements, etc.

This would be extremely valuable for teams using AI-assisted development workflows, enabling natural language queries against ELM data without writing custom scripts.

Proposed Implementation

A new module elmclient/mcp_server.py (or elmclient/examples/mcp_server.py) that:

1. Uses FastMCP from the mcp package to expose tools:

   • list_projects(domain) — List accessible projects (ccm/rm/qm)
   • list_workitems(project, query, pagesize) — Query work items via OSLC
   • get_workitem(project, id) — Get work item details
   • list_requirements(project, module) — Query DOORS Next requirements
   • get_requirement(project, id) — Get requirement details

2. Handles authentication via a credentials file or environment variables

3. Adds a console_scripts entry point: elm-mcp-server=elmclient.mcp_server:main

Dependencies

• mcp[cli] (MIT license, maintained by Anthropic)

Prior Art

We have a working proof-of-concept MCP server using elmclient that successfully connects AI assistants to our ELM instance. Happy to contribute a PR with the implementation.

References

• MCP Specification
• MCP Python SDK