## Summary Working with merge requests through AI agents used to require chaining together a sprawling set of disconnected tools just to read, search, create, or review a single MR — adding unnecessary complexity and token overhead to every interaction. Now you get a consolidated, purpose-built set of MCP tools for the entire merge request lifecycle, so your AI agents can seamlessly list, inspect, create, update, and review merge requests with far less friction and far more precision. ## Desired Outcome A comprehensive, consolidated set of MCP tools for the merge request domain. - `get_merge_request` — read a single MR; facet reader with `include:` (`diffs`, `commits`, `notes`, `pipelines`, `conflicts`, `discussions`), a `detail` knob on the `diffs` facet, and facet-scoped pagination - `list_merge_requests` — list MRs (collection reader; filter by author, assignee, reviewer, state, labels, milestone, search) - `save_merge_request` — create/update an MR (`merge_request_iid` presence selects the operation: absent → create, present → update; no `method` param) - `save_merge_request_review` — review/discussion mutations (`method` enum: reply to discussion, resolve discussion, create note, create diff note, submit review, post Duo review). Intentional second writer on MRs (review vs. field mutation)