What information does get_audit_log show?

Each entry includes the timestamp of the call, the name of the LMCP tool that was invoked, and the call source — whether it came from a local desktop client over stdio or a web AI through the Cloud Relay.

Is my audit log sent to the cloud?

No. The audit log is stored locally on your Mac. get_audit_log reads it on-device and returns it straight to your AI client. Nothing is uploaded to or kept on an LMCP server.

Can I tell which AI client made a tool call?

Yes. get_audit_log records the call source, so you can distinguish between desktop clients (like Claude Desktop or Cursor over local stdio) and web AIs (like ChatGPT or Perplexity) that connect through the Cloud Relay.

View AI Tool Call History on Mac with get_audit

The get_audit_log tool returns the recent history of LMCP tool calls made on your Mac. Every entry shows the timestamp, the tool name that was invoked, and the call source (local desktop app or cloud relay). It's a read-only window into exactly what your AI assistant has been doing on your machine — useful for transparency, debugging, and peace of mind.

A natural prompt to try: "Show me the last 20 tools my AI assistant ran and where they came from." The AI calls get_audit_log and returns a clean timeline — for example, that it read 3 emails at 9:14am from your local Claude Desktop, then searched your calendar at 9:15am via the cloud relay.

LMCP is a free, native macOS MCP server that lets AI assistants use your real Mac apps locally — no API keys, no cloud account, and nothing stored on a remote server. Download LMCP to get started with 150+ tools.

Which AI agents work?

get_audit_log works with every major AI client that speaks MCP:

Desktop clients — Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed auto-configure through a local stdio connection. No setup beyond installing LMCP.
Web AIs — ChatGPT, Claude.ai (web), Grok and Perplexity connect through the LMCP Cloud Relay connector, which securely tunnels requests to the server running on your Mac.

Whether the call came in over local stdio or the cloud relay, get_audit_log records the source so you can always tell desktop traffic apart from web-AI traffic.

Automation

get_audit_log shines when chained with other tools. After a long AI session that touched Mail, Calendar and Reminders, you can ask the assistant to pull the audit log and summarize everything it did — then have it feed that summary into a Note (create_note) or a daily brief (daily_brief). You can also use it as a safety check: "Before you continue, list the tools you've run so far," letting you confirm nothing unexpected happened before approving the next destructive action like sending an email or deleting an event.

Context

Because LMCP runs on your actual Mac, get_audit_log reflects your real activity — not a sandbox or a synthetic log. The history is the genuine record of which of your local apps (Mail, Messages, Notes, Calendar, Contacts, Safari, Finder and more) the AI touched, when, and through which connection. That real context makes the audit log trustworthy evidence rather than a guess.

Productivity

Instead of digging through scattered app logs or trying to remember what an AI did across a multi-step task, get_audit_log gives you the full picture in one prompt. It turns a tedious manual reconstruction into a two-second question. Power users rely on it to debug why an automation behaved unexpectedly, to verify a workflow ran in the right order, and to keep a quick mental audit trail of AI activity without leaving the chat.

Privacy & GDPR

The audit log lives entirely on your Mac. get_audit_log reads it locally and returns it directly to your AI client — no copy is sent to or stored on any LMCP server. Your tool-call history never leaves your machine. This local-first design makes LMCP GDPR-compliant by architecture: there is no cloud database of your activity to leak, subpoena, or breach. You own the record, and you can clear it anytime by managing the local log.