What does create_referral_invites actually return?

For each recipient you picked, it records the invite and returns a unique referral link tied to your account. You can then share those links by email, message, or however you like.

Do I need an API key or a cloud account?

No. LMCP is a free, native macOS MCP server. It runs entirely on your Mac with no API keys and stores nothing on a remote server. Web AIs connect through the LMCP Cloud Relay, but your data never leaves your machine.

Can I invite several people at once?

Yes. Pass all the recipients you selected and create_referral_invites records each one and returns a separate, unique link per person so every invite is properly attributed.

Create Referral Invites with AI on Mac

The create_referral_invites tool is step 2 of the colleague-invite flow in LMCP. Once you've picked the people you want to invite, this tool records each invite and returns a unique referral link you can share. It's the moment the loose list of names becomes real, trackable invitations — each with its own link tied back to you.

A typical prompt looks like this: "Create referral invites for Ana and Tom and give me the links to send them." Your AI assistant calls create_referral_invites with the recipients you selected, the tool registers each one, and you get back the links ready to drop into an email or a message.

Which AI agents work?

LMCP exposes 150+ tools to any MCP-capable AI assistant. The following all work with create_referral_invites:

Claude Desktop — auto-configures over local stdio
Cursor — auto-configures over local stdio
VS Code (GitHub Copilot) — auto-configures over local stdio
Windsurf — auto-configures over local stdio
Zed — auto-configures over local stdio
Claude.ai (web) — connects via the LMCP Cloud Relay connector
ChatGPT — connects via the LMCP Cloud Relay connector
Grok — connects via the LMCP Cloud Relay connector
Perplexity — connects via the LMCP Cloud Relay connector

Desktop clients talk to LMCP through a local stdio connection that sets itself up automatically. Web AIs reach the same tools securely through the LMCP Cloud Relay.

Automation

create_referral_invites is designed to chain. The natural pipeline starts with picking recipients — your AI can pull candidates from Contacts, recent Mail threads, or Messages, let you confirm who to include, and then hand that list to create_referral_invites. Once the links come back, the AI can compose a personalized email per recipient and queue it with the Mail tools, or send a quick note over Messages. One conversation goes from "who should I invite?" to "invites sent with tracking links" without you leaving the chat.

Context

Because LMCP runs on your Mac and reads your real apps, the AI works with genuine context — not placeholders. It can suggest colleagues from your actual address book, dedupe people you've already invited, and tailor the outreach message to how you normally write. The referral links it returns are real, unique, and tied to your account, so every invite is properly attributed.

Productivity

Manually generating a referral link, copying it, and pasting it per person is tedious and error-prone — easy to send the same link twice or lose track of who got what. Letting your AI run create_referral_invites turns a five-minute, multi-tab chore into a single sentence. Invite three colleagues or thirty: each gets a distinct link, recorded automatically, ready to send. That's minutes saved every time and a clean record of who you reached.

Privacy & GDPR

Everything runs locally on your Mac. LMCP needs no API keys, stores nothing on a remote server, and never copies your contacts or invite data to the cloud. The recipients you pick and the invites you record stay on your machine; only the referral link itself is shareable, by your choice. This local-first design makes LMCP GDPR-compliant by architecture. Download LMCP to get started.