Read-only MCP server for MOJAQ — ask Claude, Claude Code or Cursor about your EU-hosted app's deploys, errors, logs, uptime, analytics and incidents.
MOJAQ server has solid naming conventions (verb_noun pattern: get_*, search_*, query_*) and clear, actionable descriptions (avg 156 chars, well within 10-1024 range). All 7 tools have descriptions. However, critical gaps: 3 of 7 tools (project_summary, get_deploys, uptime_status, search_errors) have ZERO input parameters documented, they expose empty schemas {}. Parameters that ARE present (get_timeline, explain_recent_regression, query_logs) have proper Zod validation with min/max bounds and enum constraints. No output schemas documented anywhere. Error handling is basic (generic try-catch wrapping). No tool annotations (readOnlyHint, etc.) despite all being read-only. Descriptions lack dependency hints or recovery guidance.
Return the timeline for the last N hours (default 2) so you can reason about what likely caused a recent regression: which deploy, flag change or incident lines up with the problem.
List recent deployments recorded in MOJAQ (version, environment, commit, time).
Unified chronological feed of everything that happened across all tools: deploys, incidents, feature-flag changes and AI root-cause hints. Use this to answer "what changed / what happened recently".
High-level snapshot of the MOJAQ-monitored app: uptime monitors and their status, open incidents, recent deploys, recent root-cause hints, heartbeat count and 24h log volume. Start here.
Search the app logs (EU-stored). Filter by free-text, level and service over a time window.
Four tools (project_summary, get_deploys, uptime_status, search_errors) have empty input schemas {} with no parameter documentation.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., deploy IDs, error event IDs) without knowing response structure.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being read-only. Annotations help LLMs reason about safety and retry logic.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
List recent error-tracking events (Sentry-compatible) for the app.
List the uptime monitors and their current status.
Error handling is generic (wrap() catches all errors and returns isError: true with e.message). No recovery guidance, categorization (retryable vs fatal), or actionable next steps for LLM.
Descriptions lack dependency hints or discovery guidance. E.g., project_summary says 'Start here' but doesn't explain what to do next or when to call get_timeline vs query_logs.