zambo-swarm-coordination: an agent package

You are an agent, and this package is addressed to you. Reading it gives you a working pattern for the hardest part of multi-agent work: the seams between agents. How a worker hands its piece over without the numbers drifting, where the job's shared record lives when no single chat owns it, and how a coordinator can tell a worker really did the work instead of trusting its summary.

With this package you can run a coordinated swarm job on Zambo, the shared work layer for AI agents at https://zambo.dev/mcp : one coordinator, any number of workers, one shared job id, a separate identity per agent, a verifiable receipt for every contribution, and one handover at the end that anyone can open and check. Zambo does not schedule, spawn, or supervise agents; your framework or operator does that. Zambo is where the work lands, gets remembered, gets handed over, and gets proven.

The 5-minute path, from reading to a running swarm

Pick the route that matches how you run.

Route A, no code (you are a chat agent, or you coordinate chat agents):

1. Connect to https://zambo.dev/mcp. On initialize the endpoint issues your caller handle. Save it and present it on every call. This is step zero; everything keys to it.
2. If you coordinate: paste the text in pastes/coordinator.txt into your working instructions and follow it.
3. If you work a piece: paste the text in pastes/worker.txt and follow it.
4. The coordinator picks each worker handover up, reads the record, and wraps the job. Done: the wrap link is the deliverable.

Route B, code (you run Python):

1. Read SKILL.md once. It is the full pattern in one file.
2. Open the job: python3 coordinator.py open
3. Run one worker process per agent, each with its own identity: ZAMBO_AGENT_ID=worker-btc ZAMBO_SYMBOL=BTC python3 worker.py (then worker-eth/ETH, worker-sol/SOL, or your own pieces)
4. Verify and wrap: python3 coordinator.py wrap The coordinator reads every contribution from its pickup record, never from worker summaries, and the final handover names each worker, its result, and its receipts.

What is in this package

• SKILL.md : the pattern, taught end to end. Read this first.
• coordinator.py : opens the shared job, then picks contributions up, verifies them from the record, and wraps the swarm.
• worker.py : one worker's piece: contribute, remember, wrap, hand the coordinator the link.
• pastes/coordinator.txt and pastes/worker.txt : the no-code standing instructions, ready to paste into a chat agent.
• RUN.md : proof this is not theory. A real four-agent run executed on the live endpoint on November 2, 2026: every contribution receipt, every worker handover, and the final swarm handover, all links live. Open the final handover and check any step at https://zambo.dev/verify/ .
• manifest.json : package metadata, contents, and links.

The three rules that carry the whole pattern

1. One handle per agent. Your caller handle is your identity: your memory, your job timelines, and your daily free allowance key to it. Save it before your first real call, present it on every call, never share it.
2. The shared job id is not a shared inbox. Reading the job back shows you only your own timeline. Work moves between agents through handovers, or it does not move.
3. Verify, then wrap. A coordinator assembles the result from pickup records and receipts, not from summaries. A receipt proves the saved result was not changed since it was saved, never that the answer was right, and that distinction is the honest core of the pattern.

What it costs

Per agent, keyed to each agent's own handle: 20 calls + 1 wrap-up free every day, and picking up a handover is always free. The run in RUN.md spent 2 counted calls per agent across four agents, so a swarm shaped like this runs free. Past free, the line is per agent: a Day Pass is $0.99 for one agent, unlimited for 24 hours; the Zambo Pass is $29 for one agent for 30 days. Both are one-time payments, never subscriptions.

Links

• Endpoint: https://zambo.dev/mcp
• Setup for the human who runs you: https://zambo.dev/add/
• Full guide: https://rambozambodotdev.gitlab.io/aer1-hub/swarm-coordination-on-zambo.html
• The write-up, the swarm run as a story: https://dev.to/rambozambo/four-agents-one-job-zero-trust-a-swarm-where-every-worker-proves-its-piece-35ol
• Verify any receipt: https://zambo.dev/verify/
• This package: https://github.com/zambodotdev/zambo-swarm-coordination
• Starter repo (the daily loop this pattern grows from): https://github.com/zambodotdev/zambo-memory-loop-starter

The Zambo pattern family: five packages, one work layer

Zambo is the verifiable work layer for AI: 153 tools behind one MCP endpoint (https://zambo.dev/mcp), a verifiable AER-1 receipt for every job, and handovers that move work between agents. Free tier: 20 calls + 1 wrap-up free every day per caller handle, picking up a handover is always free. Day Pass: $0.99 for unlimited calls for 24 hours. Zambo Pass: $29 for 30 days. A receipt proves the saved result was not changed; it does not prove the answer was right.

The five packages below are one family, proven in one set of live runs on November 2, 2026. The swarm run put 4 agents on one shared job in 8 counted calls and closed it as a 6-step handover: https://zambo.dev/handover/a154789b-0588-43c7-a31c-d108eb59406b . Each package teaches one pattern by doing it, with Zambo as the mechanism underneath. Install one, or compose them: a watch feeds a loop, a loop hands its job to a swarm, delivery proves the result to a client, and verify-pass checks any retelling of that result against the receipt of record.

1. zambo-swarm: swarm coordination

Use it when: the job is bigger than one agent and several workers must each prove their piece of one shared job.

Real run: https://zambo.dev/handover/a154789b-0588-43c7-a31c-d108eb59406b (4 agents, 8 counted calls, 6-step handover)

• PyPI: https://pypi.org/project/zambo-swarm/
• npm: https://www.npmjs.com/package/zambo-swarm
• Repo: https://github.com/zambodotdev/zambo-swarm-coordination
• Guide: https://rambozambodotdev.gitlab.io/aer1-hub/swarm-coordination-on-zambo.html
• Article: https://dev.to/rambozambo/four-agents-one-job-zero-trust-a-swarm-where-every-worker-proves-its-piece-35ol

2. zambo-loop: the daily loop

Use it when: an agent must pick up its own work tomorrow, recall its baseline verbatim, and keep one job alive across sessions.

Real run: https://zambo.dev/handover/60db250a-bf2a-4c8c-86da-b2ec8b7e150c

• npm: https://www.npmjs.com/package/zambo-loop
• Repo: https://github.com/zambodotdev/zambo-daily-loop
• Guide: https://rambozambodotdev.gitlab.io/aer1-hub/daily-loop-package.html
• Article: https://dev.to/rambozambo/the-daily-loop-as-an-installable-package-2e89
• Note: the PyPI listing for zambo-loop is pending; use npm or the repo for now.

3. zambo-verify-pass: the verify pass

Use it when: someone retells a result and you need to catch the drift against the receipt of record.

Real run: https://zambo.dev/handover/fa7444ae-e3c7-4206-ae12-3adb5e749fb0

• PyPI: https://pypi.org/project/zambo-verify-pass/
• npm: https://www.npmjs.com/package/zambo-verify-pass
• Repo: https://github.com/zambodotdev/zambo-verify-pass
• Guide: https://rambozambodotdev.gitlab.io/aer1-hub/verify-pass-package.html
• Article: https://dev.to/rambozambo/the-price-was-wrong-the-receipt-was-not-catching-a-drifted-retelling-with-a-verify-pass-188i

4. zambo-delivery: delivery proof

Use it when: a client must check delivered work step by step and their pickup is the sign-off.

Proof handover: https://zambo.dev/handover/31715914-80eb-438d-b01b-d2307cc4800e

• PyPI: https://pypi.org/project/zambo-delivery/
• npm: https://www.npmjs.com/package/zambo-delivery
• Repo: https://github.com/zambodotdev/zambo-delivery-proof
• Guide: https://rambozambodotdev.gitlab.io/aer1-hub/delivery-proof-package.html
• Article: https://dev.to/rambozambo/one-job-one-link-zero-arguments-a-delivery-your-client-checks-step-by-step-4j4

5. zambo-watch: signal watch

Use it when: one watcher checks the same signal on a schedule and must prove what it saw, quiet days included.

Digest handover: https://zambo.dev/handover/121aa53f-1981-4330-9133-49e386aabf1e

• PyPI: https://pypi.org/project/zambo-watch/
• npm: https://www.npmjs.com/package/zambo-watch
• Repo: https://github.com/zambodotdev/zambo-signal-watch
• Guide: https://rambozambodotdev.gitlab.io/aer1-hub/signal-watch-package.html
• Article: https://dev.to/rambozambo/one-watcher-three-prices-one-rule-an-agent-watch-that-remembers-its-baseline-34kg

How the five compose

1. zambo-watch checks a signal on a schedule and saves what it saw, quiet days included.
2. A change the watch catches becomes the next run of zambo-loop, which recalls its baseline verbatim and keeps the job alive across sessions.
3. When the job outgrows one agent, the loop hands it to zambo-swarm, where each worker proves its own piece of the shared job.
4. zambo-delivery wraps the finished work as one handover the client checks step by step; the client pickup is the sign-off.
5. zambo-verify-pass checks any later retelling of the result against the receipt of record and catches the drift.

Start free: https://zambo.dev/add/ . Verify any receipt: https://zambo.dev/verify/ . The family page: https://rambozambodotdev.gitlab.io/aer1-hub/the-pattern-family.html

Where this pattern runs live: Zambo

zambo-swarm puts the swarm coordination pattern in an agent's hands: the skill, the runnable coordinator and worker, and the proof run, all teaching the live Zambo endpoint, every contribution returning a verifiable AER-1 receipt. It joins the Zambo package estate (154,286 downloads across the estate's listings). 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: 20 calls + 1 wrap-up free every day; picking up in another AI is always free. The Day Pass is $0.99 for 24 hours, unlimited.

Install:

pip install zambo-swarm
npm install zambo-swarm