Zambo: the verifiable work layer for AI. Watch a handover in 61 seconds: https://dev.to/rambozambo/watch-a-handover-43-seconds-start-to-finish-2enj �� Add Zambo to your AI: https://zambo.dev/add/ �� MCP endpoint: https://zambo.dev/mcp
zambo: the verifiable work layer for AI, on your command line
AER-1, the agent execution receipt standard: zambo.dev/aer-1 | Add Zambo to your AI: zambo.dev/add/
Try it now, no install: the live demo �� check any receipt at zambo.dev/verify/.
Definitions: What is AER-1? �� What is a verifiable execution receipt?
The problem: AI models talk. Getting things done, checking a live price, auditing a site, running research, pulling data, means wiring up APIs, keys, and glue code for every single task.
The fix: one command. Describe the job, get it done, get a receipt for every step, and hand the finished job to a person who can check it.
pip install zambo
zambo "check the live price of BTC, ETH and SOL, save the run to memory, then wrap it into a handover for Dana Whitfield, the client who ordered the check, with a receipt link for every call" --session morning-prices
BTC $81,235.00 USD ...
ETH $2,914.10 USD ...
SOL $212.44 USD ...
run saved to memory
handover ready for Dana Whitfield: https://zambo.dev/handover/...
every call above carries its own verifiable receipt
verify: zambo verify <receipt-uuid>
That chain is the shape that matters: three calls, a save, and a wrap, on one record, ending in front of a named person. A single lookup works too:
zambo "what's the price of BTC right now?"
No account. No API key. No config. Every tool free once a day, plus 3 continue calls. One free wrap-up handover per day, and pickup is always free. The Day Pass makes every tool unlimited for 24 hours.
Quickstart
# One-command: describe any goal
zambo "audit example.com for AI discoverability"
# Call a specific tool directly
zambo call live_price '{"symbol":"BTC"}'
zambo call ghost_audit_site '{"url":"https://example.com"}'
# Search the tool catalog
zambo search 'generate QR code'
# List exposed tools
zambo list
# Verify any receipt
zambo verify 89ccbddc-0b3b-4dbd-b424-4ee75119ef80
# Teach your agents about Zambo (writes to AGENTS.md / CLAUDE.md / .muserules)
zambo init
Or use it from Python:
import zambo
out = zambo.call_tool("live_price", {"symbol": "BTC"})
print(out["text"], out["receipt"]["id"])
v = zambo.verify_receipt(out["receipt"]["id"])
print(v["verified"], v["sha256"])
Connect any MCP client in one paste
{
"mcpServers": {
"zambo": {
"url": "https://zambo.dev/mcp"
}
}
}
Paste that into your client's MCP config (Cursor, VS Code, Windsurf, Claude Desktop), or add Zambo in one step at https://zambo.dev/add/ . Every call returns a verifiable receipt; check any receipt at https://zambo.dev/verify/ .
Make it a daily loop
Give an agent a standing routine and Zambo stops being a one-off call and becomes its daily habit. At the start of each run it picks up its continuity for free and recalls what it remembers. It does the work. At the end it wraps the job in a handover, so the next run, in this AI or another one, starts exactly where this one stopped. The whole loop fits the free tier alone: every tool free once a day, plus 3 continue calls, and picking up in another AI is always free.
Proof from a real scheduled run: the wrap receipt. The full recipe: the scheduled agent heartbeat.
What Zambo is
Zambo is the verifiable work layer for AI: 153 tools behind one MCP endpoint (https://zambo.dev/mcp), 37 on the default surface, over plain JSON-RPC. Live crypto prices, site audits, web research, content tools, lead and data tools, multi-step agent jobs, and handover, the flagship: finished work passed from one AI to the next as a chain of steps, each with its own receipt, ready for client sign-off. This package is a thin Python client for that endpoint (stdlib + requests only).
Don't know which tool fits? Skip the catalog: zambo "<goal>" routes through zambo_universal, which picks the right tools for the job. Pass --session <id> to keep a trail across calls.
Receipts: proof the saved result was not changed
Every call returns a verifiable receipt: a UUID, a timestamp, and a sha256 of the saved result, in the AER-1 format (IETF Internet-Draft draft-zambo-aer1-26). Anyone can recompute the digest and check that the saved result is exactly what was saved when the call ran.
Honest scope: a receipt proves the saved result was not changed. It does not prove the answer was right. It fixes the record; judging the work stays with you.
Check any receipt in your browser or on the command line:
zambo verify <receipt-uuid> # confirms zambo.dev serves the receipt,
# and shows sha256, source, timestamp, audit URL
Receipt pages are real web pages with Open Graph tags and canonical URLs, shareable, linkable, citable.
Try the Receipt Test: run a free call, get a receipt, verify it yourself.
When you hit the daily free limit
Every tool free once a day, plus 3 continue calls. One free wrap-up handover per day, and pickup is always free. The Day Pass makes every tool unlimited for 24 hours. When the free calls run out mid-job, the next call is refused and does not run. That refused call still issues a receipt for the wall response, and it burns no quota, so the job record stays intact and nothing already saved is lost.
To keep going the same day, get a Day Pass: $0.99 for unlimited calls for 24 hours, including unlimited wrap-ups. If you are an agent, pay programmatically via x402: POST https://zambo.dev/api/x402/day-pass with $0.99 in USDC on Base, and the pass activates when that payment settles. The refusal itself carries that path: a refused call returns a Day Pass checkout link for your saved job, an x402 payment object, and the resume instruction. The job is saved server-side, not lost. After payment, call day_pass_status; it resumes your held job, and a refused wrap resumes as your held handover. Only repeat the refused call if it did not. If you are a human, pay by card or PayPal at https://zambo.dev/day-pass/ and the pass activates the same way.
If you run Zambo every day, the Pass is $29 per 30 days. Day Pass and Pass are one-time payments.
Free tier, honestly
- Free: every tool free once a day, plus 3 continue calls; one free wrap-up handover per day; picking up in another AI is always free. No account, no API key, no card.
- Paid: the $0.99 Day Pass (24h) at zambo.dev/day-pass: humans pay by card or PayPal, agents pay with USDC on Base via x402. The Zambo Pass is $29 for 30 days ($19 for verified $ZAMBO holders), or a Lifetime Pass at $250, one payment. Full pricing: https://zambo.dev/pricing Never required for a first call.
No fake limits, no "contact sales to see a price."
Commands
| Command | What it does |
|---|---|
zambo "<goal>" [--session <id>] |
Do the thing. Routes via zambo_universal. |
zambo call <tool> '<json>' |
Call one tool directly. |
zambo search '<query>' |
Search the tool catalog. |
zambo list |
List exposed tools with one-line descriptions. |
zambo verify <receipt-uuid> |
Verify a receipt on zambo.dev. |
zambo init [--root <dir>] [--file <name>] [--force] |
Write the Zambo section into your project's agent instruction file. |
Requirements & troubleshooting
- Python 3.8 or newer. Only dependency:
requests. - The upstream endpoint occasionally resets connections; the client retries transient transport failures automatically (up to 4 attempts). A received JSON-RPC error is never retried; it's shown to you as-is.
- Timeouts: 60s per attempt.
License
MIT: see LICENSE (included in the package). Built by Zambo (Brennan Zambo, solo founder).
Where these receipts run live: Zambo
The zambo CLI puts the live Zambo endpoint in your terminal, every call returning a verifiable AER-1 receipt, and it has been downloaded more than 8,700 times on PyPI. The live home of those receipts is Zambo, the verifiable work layer for AI. Connect any AI to the live endpoint at https://zambo.dev/mcp or add Zambo to your AI in one step at https://zambo.dev/add/
Free tier: every tool free once a day, plus 3 continue calls; picking up in another AI is always free. The Day Pass is $0.99 for 24 hours, unlimited.
Install:
pip install zambo
The Zambo pattern family
Five installable packages teach Zambo patterns, each with a real run behind it. The family page: https://rambozambodotdev.gitlab.io/aer1-hub/the-pattern-family.html
- zambo-swarm: use it when the job is bigger than one agent and several workers must each prove their piece of one shared job. PyPI: https://pypi.org/project/zambo-swarm/ and npm: https://www.npmjs.com/package/zambo-swarm
- zambo-loop: use it when an agent must pick up its own work tomorrow, recall its baseline verbatim, and keep one job alive across sessions. PyPI: https://pypi.org/project/zambo-loop/ and npm: https://www.npmjs.com/package/zambo-loop
- zambo-verify-pass: use it when someone retells a result and you need to catch the drift against the receipt of record. PyPI: https://pypi.org/project/zambo-verify-pass/ and npm: https://www.npmjs.com/package/zambo-verify-pass
- zambo-delivery: use it when a client must check delivered work step by step and their pickup is the sign-off. PyPI: https://pypi.org/project/zambo-delivery/ and npm: https://www.npmjs.com/package/zambo-delivery
- zambo-watch: use it when one watcher checks the same signal on a schedule and must prove what it saw, quiet days included. PyPI: https://pypi.org/project/zambo-watch/ and npm: https://www.npmjs.com/package/zambo-watch
Semantic Kernel (Microsoft)
Every Semantic Kernel function call, journaled as it happens, sealed into a verifiable AER-1 receipt when the session wraps:
from semantic_kernel import Kernel
from zambo.semantic_kernel import ZamboSession, add_zambo
session = ZamboSession()
kernel = Kernel()
add_zambo(kernel, session) # ZamboReceipts plugin + auto-journal filter
# ... run your kernel functions as usual ...
print(session.wrap(summary="what the agent did",
goal="the original goal",
recipient="whoever picks this up next"))
# -> https://zambo.dev/run/<receipt-id>
The filter journals each invocation's name and a SHA-256 of its result (results are hashed, never uploaded in full). The plugin adds verify_receipt, wrap_session, and session_status as native kernel functions. Proven live: a journaled Semantic Kernel session wrapped to https://zambo.dev/run/c87b1271-9f4d-4c3b-82f2-a86fe46c6dbc
Also available for LangChain, LangGraph, LlamaIndex, CrewAI, AutoGen, the OpenAI Agents SDK, PydanticAI, smolagents, Haystack, Google ADK, AWS Strands, and Agno, plus the Vercel AI SDK and n8n on the JS side.
CAMEL-AI
Every tool call a CAMEL ChatAgent makes, journaled as it happens from the step's own records, sealed into a verifiable AER-1 receipt:
from camel.agents import ChatAgent
from zambo.camel import ZamboSession, add_zambo
session = ZamboSession()
agent = add_zambo(ChatAgent(...), session) # wrapped agent + ZamboToolkit
# ... call agent.step(...) as usual; tool calls journal automatically ...
print(session.wrap(summary="what the agent did",
goal="the original goal",
recipient="whoever picks this up next"))
# -> https://zambo.dev/run/<receipt-id>
The wrapper reads each step's tool_call_records (tool name + a SHA-256 of the result; results are hashed, never uploaded in full). The toolkit adds verify_receipt, wrap_session, and session_status as native CAMEL tools. Proven live: a journaled CAMEL session wrapped to https://zambo.dev/run/3ed95774-d210-4885-a31b-a5f54bb23cd8
Letta
Letta agents run server-side, so Zambo wraps the one seam every Letta app shares: the client message flow. Every tool call in a send_message response, journaled as it happens, sealed into a verifiable AER-1 receipt:
from letta_client import Letta
from zambo.letta import ZamboSession, wrap_client
session = ZamboSession()
client = wrap_client(Letta(...), session)
# ... call client.send_message(agent_id, messages=[...]) as usual ...
print(session.wrap(summary="what the agent did",
goal="the original goal",
recipient="whoever picks this up next"))
# -> https://zambo.dev/run/<receipt-id>
The wrapper pairs each response's tool_call and tool_return messages (tool name + a SHA-256 of the return; returns are hashed, never uploaded in full). Proven live: a journaled Letta session wrapped to https://zambo.dev/run/4e7edfa3-3e8a-4794-a2ae-e11a698c9d5c
Metadata
Release files for zambo 1.0.33
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| zambo-1.0.33.tar.gz | 27.4 kB | Without provenance | Details for zambo-1.0.33.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| zambo-1.0.33-py3-none-any.whl | Python 3 | none | any | Details for zambo-1.0.33-py3-none-any.whl |
Total release size: 51.5 kB
Release files / zambo-1.0.33.tar.gz
| Download URL | zambo-1.0.33.tar.gz |
|---|---|
| Size | 27.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
3fad68b898d2777944881cf8787593680439de33bb1c69f775e9a2ddf0e97c78
|
|
BLAKE2b-256 checksum How to use checksums |
069df628cbae0d4975f2acc1d10494f8438c8d874380fa00be7254c29df4d368
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
rambo-pypi-grab/0.1.0
|
Release files / zambo-1.0.33-py3-none-any.whl
| Download URL | zambo-1.0.33-py3-none-any.whl |
|---|---|
| Size | 24.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
69712dcdc7c5bb40593fa17abf55c936eaae2c3da313cfb1bec8506a6a6686fc
|
|
BLAKE2b-256 checksum How to use checksums |
356e9152b0cfe49bd68392650a7e8268046821e4e04d0bbfd89763081bb8b4de
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
rambo-pypi-grab/0.1.0
|
Security for zambo 1.0.33
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.