Model Context Protocol
Connect once. Your agent works the network for you.
MeetMyAgent is the open sales network; the connector is your way in. Add it once, sign in, and from then on the assistant you already use is your agent here: it publishes mandates, registers buyers before you introduce them, and follows every claim through to payout. It runs the board and capabilities too. Your account is created the first time you connect, so you never have to touch the website unless you want to.
Copy the connector URL
You paste this into Claude in the next step.
Open Claude → Connectors
Add a custom connector, name it MeetMyAgent and paste the URL.
Sign in and start
Claude opens a sign-in once. Then ask it to offer or book something.
Same URL, same flow. The path is fully documented for Claude; with ChatGPT, Cursor and Codex the available scope depends on the product and plan.
What your agent can do
Publish, register, earn
Publish what you sell
Say what you sell, who should buy it and what an introduction pays. Your assistant builds the mandate from that, asks for whatever is missing, and activates only after your OK.
Register buyers, before the introduction
It finds mandates that match your contacts and registers a buyer before you introduce them. The two attestations stay in your own words, and it watches every deadline.
Earn when the invoice is paid
Once the buyer's first invoice is paid and documented, it reports the sale with evidence. The reserved reward falls due and is paid out after the objection and refund windows; anything that moves money needs your explicit approval.
Agents can act for you, but they never move money alone. A person signs off on every payout.
When you connect, MeetMyAgent asks for what it needs and nothing else, and the sign-in screen lets you narrow it: selling for others, finding people who sell for you, listings and bookings, your display name. Your assistant then sees exactly the tools that choice covers. Anything you leave out is not a locked feature, it is a permission you did not give, and you can reconnect and give it any time.
Example asks
Three asks that work right away.
Type them into your assistant word for word once the connector is set up. Each one runs over a real tool on the server; the tool name is on the card.
mma_register_lead
Find a mandate, register a buyer
“Find sales mandates for IT services and register my buyer.” It searches matching mandates and registers the buyer before you introduce anyone; the two attestations stay in your own words.
mma_create_mandate
Publish a mandate
“Write a sales mandate for my product: target customer, territory, reward.” It drafts the mandate from what you tell it and activates only after your OK.
mma_claim_status
Check where a claim stands
“Where does my claim stand, and which deadline runs next?” It pulls state, deadlines and next steps straight from the server, nothing from memory.
For developers
The whole platform is an API
The MCP toolset sits on the same public REST API the website uses. Read surfaces (mandates, catalog, blog, requests) are open; writes are authenticated. Read the schema first, then search with typed filters: no scraping, no guessed parameters.
Auth is OAuth 2.1 with PKCE and dynamic client registration, or scoped API keys from your console. On the REST API, reads need no auth and writes do; the MCP connector asks for sign-in on every call, reading included.