Troubleshoot AI, MCP, Tunnel and Feishu
Your goal
Identify the failing layer before changing accounts or permissions.
Before you start
Record connection type, version, redacted errors and recent updates or credential rotation.
Steps
- AI Chat: verify the CLI runs and is signed in; restart after PATH changes.
- Local MCP: check endpoint, dedicated credential and its read grant; refresh tools after updates.
- Tunnel: check ID, Runtime App Key and channel state, with only one managed instance per configuration.
- Feishu: check CLI authorization, then access to the actual document/table.
leafloom_mcpindicates local authorization;feishu_cliindicates upstream errors. - After write timeouts, uncertain results or partial success, read the actual resource before retrying.
What to expect
Client caching, local grants, channel readiness and Feishu resource errors can be handled separately.
If it does not work
Keep configs/tokens private and avoid granting all writes for diagnostics. Verified account/channel status does not prove resource access.
Next steps
- Connect a local CLI and chat with AI
- Connect local MCP clients and Skills
- Connect ChatGPT through a Tunnel
- Connect Feishu and Lark