List Referral Candidates on Mac with AI

The list_referral_candidates tool is step one of recommending LMCP to a colleague. Ask your AI assistant something like "Who could I refer LMCP to, and draft me an invite?" and it returns a list of your emailable contacts pulled from your real Mac address book, alongside a playful, ready-to-send invite template. No spreadsheet, no copy-paste from Contacts.app, no guessing email addresses — the candidates and the message arrive together.

Because it reads your actual contacts locally, the suggestions are real people you already know, not a generic list. The invite template is written to be sent as-is or lightly edited, so going from "I should tell Sam about this" to a sent email takes seconds.

Download LMCP to get started — it is free and native to macOS.

Which AI assistants can use list_referral_candidates?

LMCP exposes list_referral_candidates and 230+ tools to the AI clients you already use:

• Desktop clients — Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed auto-configure over local stdio. The tools appear automatically after install.
• Web AIs — ChatGPT, Claude.ai, Grok and Perplexity connect through the LMCP Cloud Relay connector, so a browser-based assistant can still reach your Mac securely.

Whichever you prefer, the same set of tools is available once LMCP is running in your menu bar.

Can my AI automate list_referral_candidates and chain it with other apps?

This tool is the first link in a referral chain. Once it returns candidates and a template, your AI can hand the result straight to the email tools: pick a contact, fill the playful invite, and call send_email or create_draft — all in one conversation. You can ask "send the invite to my three closest design contacts" and the AI walks from candidate list to drafted, addressed messages without you switching apps.

Does list_referral_candidates work with my real Mac data?

The AI works with your real Mac data, not a sandbox. list_referral_candidates reads the contacts already on your machine — iCloud, Google, on-My-Mac — so the people it suggests are the colleagues, clients and friends you actually correspond with. That grounding is what makes the referral list useful instead of generic.

How much time does list_referral_candidates save?

Recommending a tool you like usually stalls on friction: opening Contacts, remembering who would care, finding their email, then writing something that does not sound like spam. This tool collapses that into a single prompt. What was a 10-minute detour becomes a 20-second exchange — you review the candidates, approve the playful invite, and you are done. Multiply that across the handful of people you would happily refer and it is the difference between meaning to share LMCP and actually doing it.

Is list_referral_candidates private and GDPR-compliant?

Everything runs locally on your Mac. Your contacts are read on-device to build the candidate list — there is no cloud copy, no server-side storage, and no API keys to hand over. LMCP never uploads your address book; the AI sees only what it needs to draft the invite, in the moment, on your machine. This is privacy by architecture, not by policy. Learn more in GDPR-compliant by architecture.