What does update_self_diagnosis actually return?

It returns LMCP's self-update health: the current installed version, the last several update attempts with any error messages, and whether the update cache directory is writable — so you can tell at a glance if you're up to date or stuck.

Why would an update attempt fail?

Common causes are a non-writable update cache, network interruptions during download, or permission issues. update_self_diagnosis surfaces the error message from each attempt so the AI can point you to the right fix.

Does this send my data to the cloud?

No. The tool reads your version and cache state locally on your Mac and returns it to your AI client. Nothing is uploaded or stored on an LMCP server, and no API keys are required.

Check LMCP Self-Update Health on Mac with AI

The update_self_diagnosis tool reports the health of LMCP's self-updater on your Mac. In one call it returns the current installed version, the last few update attempts (including any error messages from failed downloads or installs), and whether the update cache directory is writable. Instead of digging through logs or wondering why you're stuck on an old build, you just ask your AI assistant.

A natural prompt looks like this: "Run update_self_diagnosis and tell me if LMCP is up to date or if any updates failed." The assistant calls the tool and explains, in plain language, your version, the outcome of recent update attempts, and whether the cache is writable — so you know immediately whether a manual fix is needed.

Ready to try it? Download LMCP — it's free and native to macOS.

Which AI agents work?

update_self_diagnosis works with every major AI client LMCP supports — one of 150+ tools available the moment LMCP is installed.

Desktop clients (auto-configure via local stdio): Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf, and Zed. LMCP wires itself into their MCP config automatically.
Web AIs (connect through the LMCP Cloud Relay connector): ChatGPT, Claude.ai (web), Grok, and Perplexity. You add the LMCP connector once and the tool becomes available.

Whatever assistant you prefer, the diagnosis runs against the LMCP instance on your own Mac.

Automation

This tool is the natural first step in any update or troubleshooting workflow. Chain it with others for hands-off maintenance:

Run update_self_diagnosis first to learn your current version and whether updates are failing.
If the cache isn't writable or attempts errored, have the AI run run_diagnostics or get_config to surface permission and path issues.
Pair it with lmcp_state to confirm which services are running, then ask the AI to summarize everything into a single status report.

A prompt like "Diagnose my LMCP update health, and if anything failed, check diagnostics and tell me how to fix it" can run the whole chain at once.

Context

Because LMCP runs natively on your Mac, update_self_diagnosis reflects your actual machine — the real installed binary, the real update cache directory, and the real history of attempts. There's no guessing from a generic server-side record. The AI sees the same self-update state your Mac sees, so its advice is grounded in your specific environment rather than a one-size-fits-all answer.

Productivity

Stale versions are a silent cause of bugs and missing features. Manually checking which build you're on, whether the last auto-update succeeded, and whether a folder is writable can take ten minutes of log-spelunking. With update_self_diagnosis it's a single sentence to your AI and an answer in seconds. You catch a failed update before it costs you, and you stay on the latest, most reliable build without thinking about it.

Privacy & GDPR

Everything runs locally on your Mac. update_self_diagnosis reads your version and update cache state on-device and returns it straight to your AI client — no copy is sent to or stored on any LMCP server, and there are no API keys or cloud accounts involved. This is privacy by architecture: your data never leaves your machine. Learn more in our guide on being GDPR-compliant by architecture.