Use Talkif from Muse
Muse can browse the web, but it can’t see what happened on your calls, which voice agent answered a number, or how a campaign is going. A Muse custom connector gives it your Talkif account as tools: Muse reads and acts through the Talkif MCP server, with exactly the checks of a direct API call, and asks you before anything that costs money or reaches people.
How it fits together
Muse builds a custom connector from a message you paste into a Muse chat. The connector talks to the Talkif MCP server and sends a Talkif API key with every request. You enter the key on Muse’s own connect page, never in the chat.
The key decides what Muse can do. Muse sees only the tools whose endpoints the key’s scopes cover, and every call is checked again against the key. A key with calls:read lets Muse answer “which calls failed today?” but change nothing; add flows:* only if you want Muse to change your voice agents. To give Muse only tools that change nothing, end the endpoint with ?read_only=true.
Some tools cost money or reach people. Placing a call, starting a campaign, publishing a voice agent and the others in Tools that cost money or reach people are marked for the client to ask first. The setup message below asks Muse to get your approval for each of them, every time.
Tool results include what callers said: transcripts, names, notes. Keep Muse asking before the tools that cost money or reach people, so text in a transcript can never place a call on its own.
Connect Muse
Create an API key
In the Talkif dashboard, open Developer → Credentials → Create API Key. Name it “Muse” and choose the narrowest scopes for what you want Muse to do. Copy the key: it’s shown only once. See Creating a key.
Stop Muse’s access
Revoke the key under Developer → Credentials. Revoking is immediate: Muse’s next tool call fails with api_key_revoked. Then remove the connector in Muse.
If something goes wrong
The full list of refusals is under Errors on the MCP server page.