MCP server for syncing Backlog issues to local .tasks folder
This server has critical gaps in schema quality, parameter documentation, and error handling. While tool names follow verb-noun patterns (sync-issues, get-issue, test-connection, update-issues), the descriptions are minimal and most parameters lack detailed type information and constraints. The input schemas visible in src/server.ts show that sync-issues and test-connection use empty input objects ({}), which is problematic for validation. The get-issue tool accepts issueKey as a string but provides no format constraints (e.g., must match pattern 'PROJ-123'). The update-issues tool accepts an array of strings but lacks itemType validation and no description of what changes are expected. Error handling is present but generic ('Error syncing issues: {error}'), offering no recovery guidance. No output schemas are documented. The codebase shows basic validation but no comprehensive error classification or user-actionable recovery paths.
Get details of a specific Backlog issue
Sync issues from Backlog to local .tasks folder
Test connection to Backlog API
Update Backlog issues with changes from local task files
sync-issues and test-connection have empty input schemas ({}). These tools accept no parameters, which is valid, but the absence of explicit validation or documentation about prerequisites (apiKey, baseUrl, projectKey must be configured server-side) means LLMs cannot determine if the tool is ready to call.
get-issue parameter 'issueKey' has a description 'Issue key (e.g., PROJ-123)' that embeds an example value. Should replace with: 'Issue key matching pattern UPPERCASE-NUMBER (e.g., PROJ-123 format)' or use a regex constraint.
update-issues input schema defines issueKeys as array of strings, but provides no itemType constraint or format documentation. No description of what 'changes from local task files' means or what fields are updatable. LLMs cannot determine valid update payloads.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 37 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 17 | - | v1 |
No output schemas documented for any tool. LLMs cannot predict what fields to expect or plan downstream tool calls. sync-issues returns a text summary; get-issue returns stringified JSON; update-issues returns unknown format. Violates response-shaper pattern.
Error handling is generic. All tools return 'Error syncing issues: {error}' or similar raw error strings. No actionable recovery guidance (e.g., 'Connection failed. Verify apiKey and baseUrl in server config.' or 'Issue not found. Call search_issues() to find the correct key.'). Violates recovery-guide pattern.
Tool descriptions are too brief (50 chars avg). 'Sync issues from Backlog to local .tasks folder' lacks context on WHEN to use, WHAT happens, and prerequisites. 'Get details of a specific Backlog issue' omits what fields are returned. Baseline is 194 chars; these descriptions are <25% of target. Descriptions under 50 chars cannot effectively guide LLM tool selection.
sync-issues modifies state (writes to .tasks folder) but description does not explicitly state this is a WRITE operation. Agents need to know which calls are safe to retry vs. idempotent. sync-issues can be called multiple times; update-issues irreversibly changes issues in Backlog. This distinction is not documented.
No pagination or result limits documented. sync-issues returns all issues matching a query. If a Backlog project has hundreds or thousands of issues, this could exhaust token budgets or cause timeouts. No page/limit parameters or documented cap on returned items.
Credentials (apiKey, baseUrl, projectKey) are stored in server-side config, which is correct. However, the tool descriptions do not clarify that these are required and must be configured server-side. An LLM might attempt to pass them as parameters, which would fail and cause confusion.