How do I add Zambo to Muse?

Ask Muse, in any chat, to create a custom connector for Zambo: give it the name Zambo, the MCP endpoint https://zambo.dev/mcp, the transport (remote streamable HTTP), and the auth (none). Muse builds the connector itself, connects, and tests the tools. There is no settings form for adding a server, nothing to install, and no API key to paste, because Zambo's free tier needs none. You can review or disconnect the connector any time under Settings, then Connectors.

What do I paste into Muse?

https://zambo.dev/mcp. Paste that URL in one message asking Muse to create a custom connector for Zambo, a remote MCP server (transport: remote streamable HTTP; auth: none, no API key). That single URL is the only thing Muse needs from you; it builds the rest itself.

Can a Muse agent leave receipts for its work?

Yes, for the work it does through Zambo's tools. When your Muse agent calls a Zambo tool, the executing server captures the record at the moment of the call: the tool, a hash of the arguments, the exact result bytes committed under SHA-256, and a timestamp. The answer comes back with a receipt at a public URL. Work your Muse agent describes only in its own summary, with no tool call behind it, leaves no receipt; the receipt covers what actually ran, which is the part worth proving.

How do I verify work my Muse agent did?

Open the receipt URL from the chat. The page shows what ran, what it was given, and the exact saved result. Check it at zambo.dev/verify/, or recompute the SHA-256 commitments yourself from the bytes on the page; nothing about the check requires trusting Zambo or Muse. Because a receipt is a portable record in the open AER-1 format, you can also hand the URL to a different AI, in ChatGPT, Grok, or Cursor, and that session can fetch and recompute the same record.

Does Zambo work with the Muse connector directory?

Yes, in two ways. Today, the custom connector path works with any Muse account: ask Muse in a chat to create a custom connector for https://zambo.dev/mcp, and Zambo's tools are available in your chats, no listing required. Separately, Zambo's own submission to the Muse connector directory is currently in review. When that review completes, adding Zambo becomes a one-tap pick from the directory itself; until then, the custom connector is the same server, the same receipts, one paste away.

A Receipt for This Conversation: One Zambo Session, Checked End to End

The problem

The invoice was written in BTC terms. The client says the rate used was wrong. What settles it is not a better argument; it is the price the tool actually returned at the time, saved in a record nobody quietly edits afterward.

Call 1: find the right tool

capability_search, query "check the price of a cryptocurrency". The session asked Zambo which tool fits before calling anything. Zambo returned an execution plan naming live_price at rank 1.

Receipt: https://zambo.dev/run/d81aeb41-9954-4367-9ba1-1fbf466eef90

Call 2: the price, with its source

live_price, symbol BTC. Returned: BTC $79,544.00 USD, down 2.36% over 24 hours, market cap $1599.85B, live via CoinGecko, Sat, 24 Oct 2026 16:25:38 GMT.

Receipt: https://zambo.dev/run/26b353a9-8e34-4820-a351-9a65208e11c1

Call 3: write it into the record

journal_log. The finding was written into job ws6-muse-demo-20261024 as a chained journal entry, so the answer to the client is a record, not a recollection.

Receipt: https://zambo.dev/run/1f56fe3e-b4f3-40b0-8619-5935741c9a73

Check it yourself

Open any receipt link above; each shows the tool, the time, and the saved result with its sha-256 digest. The journal chain check for the whole job is public: https://zambo.dev/api/journal/ws6-muse-demo-20261024/verify returns valid. An agent receipt is a record of one ai agent tool call that anyone can check: it keeps the exact saved bytes and their sha-256 digest, and it proves the result was not changed since it was saved, not that it was right.

What this proves, and what it does not

The receipts prove the saved results were not changed after they were saved, and the journal chain proves the order the entries were written in. They do not prove the price was right; CoinGecko is the source named in the record, and the record is what lets the client check the same source. Proof of the record, not proof of the world.

If the conversation itself is later deleted, the receipt links are what remain checkable: how can I check AI work if the chat has been deleted? The record outlives the chat because it was saved when the calls ran.

Run the same session from Muse

Ask Muse: "Build a custom connector to Zambo. Name: Zambo. MCP server URL: https://zambo.dev/mcp. Transport: remote streamable HTTP. Auth: none. Test it with a first call and show me the receipt link it returns." The custom connector route works today; Zambo's Muse directory listing is in review with Meta. 20 free calls per day, then the $0.99 Day Pass.

Full Muse guide: Use Zambo with Muse. The proof pattern in general: proof of execution for AI agents. Check any receipt: verify AI work.

Frequently asked questions

Can I get a receipt for a conversation with my AI agent?

Yes, when the work runs through Zambo's tools. Every call returns a verifiable receipt at a public link (zambo.dev/run/...), and calls in one job chain together, so the whole session, not just one answer, can be checked afterward.

What does the receipt prove?

An agent receipt is a record of one ai agent tool call that anyone can check: it keeps the exact saved bytes and their sha-256 digest, and it proves the result was not changed since it was saved, not that it was right.

an agent receipt is a record of one ai agent tool call that anyone can check: it keeps the exact saved bytes and their sha-256 digest, and it proves the result was not changed since it was saved, not that it was right.