Multi-transport MCP server router and gateway for Jarvis, managing distributed MCP servers (distilled workflows, email, browser, etc.) with dashboard UI, authentication, and tool introspection.
Three tools with minimal documentation. All have basic descriptions (40-60 chars, below the 194-char baseline), but lack parameter descriptions entirely. No input schemas visible in source, only parameter names and types inferred from brief JSON snippets. No output schemas documented. Tools perform sensitive operations (authentication, restart, logout) without error handling guidance or recovery patterns. Naming is verb-based but descriptions are too terse to guide LLM selection. No evidence of idempotency, confirmation patterns, or permission gates for destructive operations.
Authenticate an MCP server by name, triggering OAuth flow or API key setup
Disconnect an MCP server by clearing stored OAuth credentials
Restart an MCP server by name, killing existing processes and restarting the daemon
No input parameter descriptions. The 'name' parameter in all three tools lacks explanation of what it accepts (server name? hostname? identifier format?), forcing LLMs to guess.
No output schemas documented. LLMs cannot plan downstream actions or extract success/failure signals. What does authenticate return? A token? A status? Unknown.
Destructive operations (restart, logout) lack confirmation or dry-run patterns. No error handling guidance. LLMs cannot distinguish retryable from fatal failures.
Tool descriptions are 40 - 60 characters (well below 194-char baseline). Too terse to explain when to use authenticate vs logout, or what 'restart' does to running processes.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 33 | 2026-07-28+ | v2 |
No permission gates or audit trails visible. Sensitive operations (restart daemon, clear credentials) should verify caller authority and log actions for compliance.