What does run_qa actually return?

It returns a QA test protocol: a structured list of every LMCP tool to test, sample parameters for each, and the result that counts as a pass. Your AI assistant then runs the listed read-only checks and reports which tools pass or fail.

Does run_qa change any of my data?

No. The protocol focuses on read-only checks, and destructive actions like sending or deleting are tested via preview only. Running it is safe and won't alter your Mail, Calendar, Contacts or Messages.

Do I need API keys or a cloud account?

No. LMCP is a free, native macOS server that runs entirely on your Mac. There are no API keys, no cloud storage, and nothing is uploaded — desktop AIs connect via local stdio and web AIs via the LMCP Cloud Relay.

Run a QA Test Protocol for Every LMCP Tool on Mac

The run_qa tool returns a ready-to-run QA test protocol covering every LMCP tool installed on your Mac. For each tool it lists what to call, sample parameters, and exactly what a passing result looks like — so you (or your AI assistant) can sweep all 150+ tools and confirm Mail, Calendar, Contacts, Notes, Reminders, Messages, Safari and the rest are actually working before you rely on them.

A natural prompt you might type: "Run the LMCP QA protocol and tell me which tools pass and which fail." The assistant calls run_qa, gets the protocol, then executes the listed read-only checks one by one and reports a green/red summary.

Ready to try it? Download LMCP and connect your AI client.

Which AI agents work?

LMCP works with every major MCP-capable assistant:

Desktop clients — Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed auto-configure through a local stdio connection. No tokens, no setup files to hand-edit.
Web AIs — ChatGPT, Claude.ai (web), Grok and Perplexity connect through the LMCP Cloud Relay connector, which securely bridges the web assistant to the tools running on your Mac.

The same run_qa protocol is available across all of them, so you can verify your setup from whichever assistant you use.

Automation

The real power of run_qa is chaining. The tool itself only returns the protocol — your AI then drives the rest. It reads the sample params, calls each tool in turn (list_emails, list_calendar_events, list_contacts, search_messages, list_reminders, and so on), compares the live output against the documented "passing result," and produces a single health report. One prompt becomes a full diagnostic sweep, with no manual clicking through each app.

Context

This is not a mock test. run_qa guides the assistant through your real Mac data — your actual mailboxes, calendars, contacts and chats. A passing run means the tools can genuinely read your live information, so when you later ask "summarize today's unread email" or "what's on my calendar tomorrow," you already know the plumbing works against the real apps on your machine.

Productivity

Manually checking that 150+ tools across two dozen apps still work after an update would take an hour of tedious poking. run_qa turns that into a 60-second AI sweep: ask once, get a categorized pass/fail list, and immediately see if, say, Messages or Teams needs a permission re-grant. It's the fastest way to confirm a fresh install or a new version is fully wired before you depend on it for real work.

Privacy & GDPR

Everything run_qa orchestrates runs locally on your Mac. The protocol is generated on-device, the test calls hit your local apps directly, and no email, calendar entry, contact or message is ever copied to a server or to LMCP. There are no API keys and nothing is stored in the cloud. This is privacy by design — see GDPR-compliant by architecture for how the local-only model keeps your data yours.