Model Context Protocol integration for Bitbucket Server - provides tools for interacting with Bitbucket repositories, branches, commits, pull requests, and files
18 tools with consistent verb-noun naming (list_, get_, create_, update_, merge_, decline_, branch_compare, pr_diff). All have descriptions (10-100 chars, mostly 50-80 chars). Input schemas visible with typed parameters and descriptions. However, output schemas are NOT documented in the tool definitions, LLMs cannot see what fields to expect from responses. Parameter descriptions lack format constraints (e.g., 'project_key' says 'e.g., PROJ' but no regex/length bounds). No error handling guidance, no recovery hints. Pagination parameters present but no total_count or next_cursor documented. WRITE and IRREVERSIBLE tools lack confirmation/dry-run patterns.
Compares commits between two branches
Creates a new branch in a repository
Creates a new pull request
Declines a pull request
Gets a specific commit by its hash
Gets the raw content of a file
Gets a specific pull request by ID
Output schemas not documented. LLMs cannot see what fields responses contain (e.g., does list_projects return total_count? is it paginated?). Forces agents to guess or fail on field extraction.
Parameter descriptions lack format constraints. 'project_key' says 'e.g., PROJ' but no length/regex bounds. LLMs may pass invalid values like 'my-project' or 'proj' without guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Gets a specific project by its key
Gets a specific repository by project key and repository slug
Lists branches in a repository with pagination
Lists commits in a repository with pagination
Lists files and directories at a given path in a repository
Lists all projects with pagination
Lists pull requests in a repository with pagination
Lists repositories in a project with pagination
Merges a pull request
Gets the diff for a pull request
Updates an existing pull request
WRITE and IRREVERSIBLE tools (create_branch, create_pr, update_pr, merge_pr, decline_pr) lack confirmation/dry-run patterns. Agents can destructively merge PRs or create branches without safeguards.
No error handling guidance. Tool descriptions do not explain what to do if a project_key is invalid, a branch already exists, or a PR cannot be merged. LLMs receive raw errors with no recovery path.
Pagination parameters (start, limit) present but no documentation of total_count, isLastPage, or next_cursor in responses. Agents cannot determine if more results exist or how to iterate safely.