If you do not write code, this is the whole product. The assistant does the calling; you read the answer.
1. Copy the address
Press Connect your AI on the dashboard. It gives you one web address with your key in it.
2. Paste it into the assistant
Wherever it asks for an MCP server. In Claude that is the connectors section of settings; other assistants put it elsewhere, but all of them want the same single address.
3. Ask for something
- What did people say on my last post?
- Draft a reply to everyone who messaged me today, and show me the drafts before sending.
- Find posts about hiring in Berlin this week and summarise who is posting them.
That address contains your key. Treat it like a password: do not paste it into a document, a chat message, or anywhere a screenshot might end up.
If the assistant cannot see the tools
Three things, in order: the address was copied whole including the key; the assistant actually supports MCP; the plan is active. The second is the one that catches people out — support arrived at different times in different products.
Questions people ask
What is the difference between this and a project room?
This is a private conversation between you and your assistant, paid for by your assistant's own subscription. A project room is shared: your team, guests, built-in specialists and your own agents in one thread, on your organization's AI key.
Can my assistant post into a project room?
Not on its own. A room accepts writes from an outside AI agent, which is something you run — a script or an agent framework — holding that room's token.