Everything else is a competent read-and-write layer over Mantis issues; the aggregates are what you cannot get in one click from the UI. On the query side, follow the server's own guidance and pass select with id, summary and description — a full issue list returns enough data to cause problems.
An MCP server over the MantisBT REST API, authenticated with a personal API token, covering issue reading and writing plus two aggregate views.
- get_issues with filters on project, status, handler and reporter plus a keyword search, paged 20 at a time, and a select list that decides which fields come back
- get_issue_by_id for a single issue in full
- The surrounding entities: the project list, a project's users, one user looked up by username, and a brute-force listing of every user
- create_issue with summary, description, project, priority, severity, handler, category and additional information
- update_issue for status, summary, priority, severity, handler, resolution and description
- add_issue_note to attach a note with its own view state
- get_issue_statistics grouped by status, priority, severity, handler or reporter, over all time, today, this week or this month
- get_assignment_statistics per user, with or without unassigned issues, optionally filtered to particular statuses
A MantisBT instance with its REST API reachable, and an API token created under My Account > API Tokens. Two variables carry it: MANTIS_API_URL pointing at the .../api/rest endpoint, and MANTIS_API_KEY. Optional settings: NODE_ENV, LOG_LEVEL, CACHE_ENABLED with CACHE_TTL_SECONDS (300 by default), and ENABLE_FILE_LOGGING with LOG_DIR. Run it over stdio with npx -y mantis-mcp-server@latest, or install globally and point the client at dist/index.js.
One command plus a key — npm install mantis-mcp-server, then supply credentials