MCP server for Jira ticket details, branch creation, and Bitbucket PR with reviewers
Server has 9 tools with reasonable naming and descriptions, but lacks structured output schemas, error handling guidance, and several parameter descriptions are generic or incomplete. Schemas are defined as Zod objects in code (good), but output structures are not documented. Most tools accept sensible parameters with decent coverage, though some lack validation constraints. No tool annotations (readOnly hints, destructive hints) present. Error responses are not addressed. Security is adequate (credentials via env vars, not params). Tool composition is reasonable but some operations could be idempotent-aware.
Approve an open Bitbucket Cloud pull request (BITBUCKET_API_TOKEN; with BITBUCKET_USER or BITBUCKET_USERNAME use app password + Basic auth). User must have permission to approve.
Assign reviewers to a Bitbucket PR by usernames. Looks up usernames and assigns to the PR (requires write access to PR).
Given a Jira browse URL or issue key, return a suggested git branch name that includes the ticket key (e.g. https://example.atlassian.net/browse/SW-144906 → feature/SW-144906). No API call; parsing only.
Create and checkout a new git branch named from a Jira ticket (e.g. feature/PROJ-123-short-description). Run from the repo root or pass repoDir.
Create a Bitbucket Cloud pull request from the current branch to the destination branch.
Fetch attachment (e.g. image) content from a Jira issue for the AI to use. Returns file buffer, MIME type, and filename.
No output schemas documented. Tools return JSON text without documented field structure, forcing LLMs to parse unstructured output. Users cannot know what fields to expect (e.g., what fields does get_jira_ticket return? attachment structure for get_jira_attachment?).
No error handling guidance. Tools do not document what errors can occur, how they should be classified (retryable vs user-fixable vs fatal), or what the LLM should do next. E.g., create_pull_request could fail if workspace/repoSlug are missing from env, no error recovery path.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 53 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 47 | - | v1 |
Fetch Jira issue/ticket details by issue key (e.g. PROJ-123). Use this to get summary, description, status, and link before creating a branch or PR.
Get reviewer names from project's mcp.json config file. Returns list of default reviewer usernames for assigning to PRs.
List Jira issues assigned to you. Uses JIRA_EMAIL as the authenticated user (assignee = currentUser()). By default returns non-done issues only, newest first.
Missing tool annotations. No readOnlyHint or destructiveHint attributes on tools. LLM cannot determine which tools are safe to retry (get_jira_ticket is read-only and idempotent; create_branch and create_pull_request are destructive and require care).
Incomplete parameter constraints. 'maxResults' in list_my_jira_issues declares min/max (1 - 100), which is good, but most other numeric parameters lack bounds. 'pullRequestId' is number but no range documented. Numeric parameters without explicit bounds let LLMs pass absurd values.
Generic parameter descriptions. 'workspace' param in create_pull_request, approve_pull_request, assign_pr_reviewers has description '(default: BITBUCKET_WORKSPACE env var)', does not explain what a workspace is, why it's needed, or valid format. Same for 'repoSlug'.
Missing chaining identifiers. get_jira_ticket likely returns issue key, description, and metadata, but it's not clear if it returns URL/link. create_pull_request could accept jiraKey to auto-format title as '[JIRA-123] Title', but requires manual formatting. No helper to chain Jira → branch → PR without extra reasoning.
No idempotence guarantees documented. create_branch could fail if branch already exists. Should tool be idempotent (return existing branch) or error? Not documented. Agents retry on failure, unclear what happens on second call with same params.
Vague parameter names. 'slug' in branch_name_for_jira and create_branch is unclear, description says 'short slug' but doesn't define what a slug is or show examples beyond 'fix-login'. A user unfamiliar with naming conventions won't understand what to pass.