MCP Tools for Pull Requests and Pipelines
AI agents can now query pull requests and CI/CD pipeline runs directly via MCP — cross-provider, filtered, and paginated. Plus: smarter empty states in the pipeline view.
Two additions for working with pull requests and pipelines: three new MCP tools that let AI agents query PR and CI data directly, and clearer empty states in the pipeline view.
New MCP Tools: query Pull Requests and Pipelines
The MCP server now exposes three new read-only tools that let AI assistants query pull request and pipeline data — without opening the dashboard.
pull_requests_list
Returns a paginated overview of all pull and merge requests across connected providers (GitHub, GitLab, Bitbucket). You can filter by:
- Provider — GitHub, GitLab, or Bitbucket
- State — Open, Draft, Merged, or Closed
- Board — single board or all
- Repository, author, branch — free-text search
- Only open or only failed — as quick filters
pull_requests_get
Returns full details of a single pull request:
- Title, branch, repository, author
- All CI pipeline steps with status and duration
- Reviews with reviewer, status, and comment
- Recent events
pipelines_list
Shows CI/CD pipeline runs grouped by status bucket:
- Needs attention (default) — failed runs, manual gates, stuck jobs
- Running, Failed, Success, or All
Filters for provider and board are available. Results include per-bucket counts, so the agent can assess overall health at a glance.
All tools at a glance
| Tool | Description |
|---|---|
pull_requests_list |
Paginated PR/MR overview, filterable by provider, state, board, repo, author, branch |
pull_requests_get |
Full PR details with CI steps, reviews, and events |
pipelines_list |
CI runs by status bucket, with counts and provider/board filter |
All three tools are read-only and scoped to the boards the connected user can access. The required scope is mcp:read — the same as boards_list and boards_get.
Pipeline view: smarter empty states
The pipeline view now shows context-aware messages when no runs are displayed.
All clear
When pipeline data exists but the active filter (default: "Needs attention") hides all of it — because everything is green — you now see a clear confirmation instead of a blank page. A button jumps directly to the unfiltered view.
No data yet
When no pipeline data exists at all, the view explains the prerequisites (active Git integration, linked repository, open PR with CI) and links directly to integration settings.
Frequently asked questions
Quick answers to the most common questions about this topic.
What new MCP tools are available for pull requests?
pull_requests_list returns a paginated, filterable overview of all pull and merge requests across GitHub, GitLab, and Bitbucket. pull_requests_get returns full details of a single PR — including CI steps, reviews, and events.What can I do with the pipeline MCP tool?
pipelines_list shows CI/CD runs grouped by status bucket (Needs Attention, Running, Failed, Success). You can filter by provider, board, and status — just like in the browser pipeline view.Do I need special permissions for the new MCP tools?
mcp:read scope and are restricted to the boards you can access. No additional permissions are required.What changed about the pipeline view empty states?
Keep reading
MCP tools for all ticket properties and per-project billability controls
AI agents can now set every ticket property and custom field over MCP. Plus: per-project billability controls and MOCO booking items named by external reference.
Preview Environments, Pipeline Dashboard, and OpenAI in the Coding Runner
New live previews directly in tickets, a cross-provider pipeline dashboard, OpenAI models in the coding runner, project templates, and new automation triggers for pull request status.
Project Folders, Help Mode, and Images in Subtasks
Projects can now be grouped into folders, a new help mode shows contextual documentation for every UI element, and subtasks support inline images. Plus: recurring todos on specific weekdays, timestamps on pull request rows, and a server-side dashboard checklist.