Your agent needs tools. It does not need another secret to leak. Here is keyless calling that still leaves proof.

Every new tool seems to start with the same chore: sign up, find the keys page, copy a secret somewhere it will probably leak from later. Zambo skips it. The MCP endpoint at https://zambo.dev/mcp takes calls with no API key at all, 153 tools behind it, 20 free calls per day. This page explains how that works, and what replaces the key for accountability: the receipt.

Try it now: add Zambo to your AI with no key step, or verify the worked example receipt from this page, also with no key.

Published October 25, 2026. The worked example below is a live receipt from the public stream, checked the same day.

Which APIs can AI agents call without an API key?

Zambo can. Its MCP endpoint at https://zambo.dev/mcp is keyless: connect a client, call a tool, get a result and a receipt, with no API key anywhere in the flow. The endpoint carries 153 tools, 35 on the default surface, with 20 free calls per day and no account or card to start. Zambo is the verifiable work layer for AI, and keyless access is deliberate, not an oversight.

Scope, stated once: 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.

Receipts follow the AER-1 format, specified by the IETF Internet-Draft draft-zambo-aer1-26. HANDOVER, Zambo's flagship product, runs on the same endpoint, so a handover packet built from keyless calls carries the same checkable record as any other Zambo work.

Why keys are the friction, and the risk

Anyone who has wired an agent to a new tool knows the ritual. Create an account. Find the keys page. Generate a secret. Paste it into a config file or a chat window. Hope it does not leak. Rotate it later and update every place it lives. Half the setup time is key handling, and every copy of the key is a copy that can end up in a log, a screenshot, a repo, or an agent's long term memory.

For an agent, the risk is sharper. Agents read files, quote context, and hand work to other agents. A secret that sits inside that context can travel with it. Removing the key removes the whole class of accident: there is nothing to paste, nothing to store, and nothing to leak.

  • Setup goes to near zero: add the endpoint at zambo.dev/add and call.
  • Sharing gets safer: you can hand an agent, a teammate, or another AI the endpoint address freely. The address is public by design.
  • Rotation disappears: no key means no rotation schedule, no stale copies, and no broken jobs after a reset.

How keyless calling works, step by step

  1. Connect: point your MCP client at https://zambo.dev/mcp. No key header, no token step.
  2. Call: ask for the outcome you need. The endpoint routes to the right tool among the 153 available, 35 of them on the default surface your client sees first.
  3. Stay inside the allowance, or pass it: the first 20 free calls per day need nothing else. Past that, a Day Pass at $0.99 USDC on Base for 24h or a Zambo Pass at $29 for 30 days, one payment, keeps the same keyless flow running.
  4. Keep the receipt: every call returns its receipt link, zambo.dev/run/ plus the receipt ID. That link is your record, and it is safe to share because it proves a result, it does not unlock anything.

No key does not mean no accountability

The usual reason for a key is accountability: who called, and can the record be trusted. Zambo answers the second half directly and the first half honestly. The receipt records the call, the tool, the time, and the exact saved result with its sha-256 digest. Anyone can recompute that digest at zambo.dev/verify and confirm the saved result was not changed since it was saved.

What the receipt does not do is just as important. It does not prove the answer was right, and it does not turn a public link into an identity check. It gives you something a key never gave you: a record of the work itself that a third party can check without asking anyone for permission.

Worked example: a keyless call, checked live

From the public stream on October 25, 2026: a capability_search call saved at 2026-10-25T18:35:28Z, receipt at zambo.dev/run/48493b66-f9bf-4bff-80cb-052af5716698, saved result digest sha256:a2b6680939153af26134fbdd5aca53dc2544f8e8a2dcd9f1122975d585f30d4d.

The verification endpoint returned verified for this receipt. Note what you needed to check it: the public link, and nothing else. No key, no login, no permission from the caller. That is the accountability model working as designed. The call needed no secret to run, and the proof needs no secret to check.

When you do need ongoing access

PlanAccessPrice
FreeKeyless calls, 20 free calls per day$0, no account, no card
Day Pass24 hours of full access, same keyless endpoint$0.99 USDC on Base for 24h
Zambo Pass30 days, one payment, annual option at $290$29 for 30 days

Paid access changes the allowance, not the shape. Same endpoint, same tools, same receipt on every call, still no API key to manage. Start free at zambo.dev/add and notice what is missing from the setup: the keys page.

Questions, answered

Which APIs can AI agents call without an API key?

Zambo is built for exactly that. Point any MCP client at https://zambo.dev/mcp and call its 153 tools with no API key, no account, and no card to start. You get 20 free calls per day, and every call returns a receipt anyone can check at zambo.dev/verify.

How does calling without a key work in practice?

Your client connects to the MCP endpoint and calls tools the normal MCP way. There is no key to create, store, rotate, or paste into a config file. The endpoint handles the free allowance and paid Pass access without a secret sitting in your project, your chat, or your agent's memory.

If there is no key, what stops abuse and what proves a call happened?

Rate limits handle the allowance, and the receipt handles the proof. Every call issues an AER-1 receipt that keeps the exact saved bytes and their sha-256 digest. Anyone with the receipt link can verify the saved result was not changed since it was saved. Accountability comes from the checkable record, not from a shared secret.

Does verified mean the result was correct?

No. Verified means the saved result was not changed since it was saved, not that it was right. The receipt lets you judge the actual output the agent produced, instead of a later edit of it. Judging quality is still review. Knowing the record is intact is what the receipt settles.