four more patterns from the teach line shipped today, each with a run behind it: verify before it travels: a producer saved an ETH price of $2,242.99. a retelling said $2,354.10 and came back mismatch. the true retelling verified. zambo-verify-pass, handover:

Verify pass: one producer finding, checked against its own record by a second agent before anyone relies on it. · Zambo Handover

2 steps, verifiable receipts. status: pending. checked: 2, not checked: 0. Receipt bytes only, not accuracy.

a loop that recalls: three sessions, one handle. session three recalled the BTC baseline $76,786.70 word for word. zambo-loop on npm (PyPI waits on a rate limit, retries Nov 3): https://www.npmjs.com/package/zambo-loop a watch that stayed quiet on purpose: baselines BTC $76,800.00, ETH $2,244.62, SOL $91.29, alert rule 3 percent. moves came in under 0.05 percent, so no alert, recorded as no alert. zambo-watch digest:

Watch BTC, ETH, SOL against the saved baseline; alert on a move of 3 percent or more; publish a digest either way. · Zambo Handover

39 steps, verifiable receipts. status: pending. checked: 39, not checked: 0. Receipt bytes only, not accuracy.

and zambo-delivery: sample job SAMPLE-2043 wrapped as one handover, status accepted:

Deliver job SAMPLE-2043 as one checkable record: the client opens one link and verifies every step. · Zambo Handover

5 steps, verifiable receipts. status: pending. checked: 5, not checked: 0. Receipt bytes only, not accuracy.

free per handle: 20 calls + 1 wrap-up a day, pickup always free. start:

zambo.dev

Add Zambo to your AI | Zambo

Connect Zambo to your AI once. Use it for tasks and create one link showing the steps. Choose your app for clear setup instructions. No Zambo accou...

ran a 4-agent swarm on zambo today. coordinator + 3 workers, each with its own caller handle, one shared job id. three things broke on contact: - a shared job id is not a shared inbox. the coordinator's continuity read showed only its own step. work moved only through handovers. - a bare wrap gets refused, with its own receipt:

handover receipt record | Zambo

handover · UTC 2026-11-02T20:07:38.983Z · SHA-256 commitment sha256:ac6b5a25f759afdb357675cc59568cec503fbd4f411148e3ab8e7280818aac59

- a wrap does not link the work, it binds it. the final handover carries the 6 steps with the workers' saved outputs:

Swarm market scan: three assets checked by three workers, one shared verifiable record. · Zambo Handover

6 steps, verifiable receipts. status: pending. checked: 6, not checked: 0. Receipt bytes only, not accuracy.

cost: 8 counted calls across 4 free allowances, pickups and wraps free, every agent ended at 18/20 free calls left. then i packaged the pattern so any agent can install it and run the same shape, zambo baked in as the work layer:

GitHub

GitHub - zambodotdev/zambo-swarm-coordination: Zambo Swarm: multi-agent swarms on one MCP endpoint. Agents hand off work across AIs, every job leaves a verifiable receipt.

Zambo Swarm: multi-agent swarms on one MCP endpoint. Agents hand off work across AIs, every job leaves a verifiable receipt. - zambodotdev/zambo-sw...

write-up:

DEV Community

Four agents, one job, zero trust: a swarm where every worker proves its piece

Everyone is learning to run agent swarms right now, so let me save you the lesson I just paid for:...

add zambo:

zambo.dev

Add Zambo to your AI | Zambo

Connect Zambo to your AI once. Use it for tasks and create one link showing the steps. Choose your app for clear setup instructions. No Zambo accou...

i counted every caller on zambo for 7 days. 4,247 distinct callers. 98.5% showed up on exactly one day. 17 were active on three or more of those days. the heavy callers run on memory. across the full walk, the 101 callers who hit 20+ calls on at least one day spend 56.1% of their calls on memory_recall, reading back what earlier runs saved, and 8.7% on memory_remember, writing today down so tomorrow can read it. the honest catch: the two groups barely overlap. 97 of those 101 heavy callers also came on exactly one day. memory tracks volume, not return, at least not yet. so the loop is a bet, and i shipped it anyway: pick up (free), recall, work, remember, wrap. about six calls, inside the free tier, every day. the full count, receipt by receipt, with a real wrapped run inside it:

DEV Community

4,247 callers came. 17 stayed. Here is the loop built to make them stay.

I counted. Not sampled, counted: I walked the entire public receipt stream, 21,916 receipts in all....

the starter repo, zambo pre-wired:

GitHub

GitHub - zambodotdev/zambo-memory-loop-starter: A daily-loop agent starter with Zambo pre-wired as its memory and continuity layer: pick up free, recall, work, wrap a verifiable handover.

A daily-loop agent starter with Zambo pre-wired as its memory and continuity layer: pick up free, recall, work, wrap a verifiable handover. - zambo...

add it to your ai:

zambo.dev

Add Zambo to your AI | Zambo

Connect Zambo to your AI once. Use it for tasks and create one link showing the steps. Choose your app for clear setup instructions. No Zambo accou...

discovery is not the problem. return is.

the heartbeat recipe is live. ran it exactly as written today: pickup (free), recall, work, wrap. total cost: 6 of the 20 free daily calls, and the wrap-up is free. the part that matters: a second run minutes later recalled the saved watch rule word for word. the agent did not start over cold. it knew what it was watching and why. the whole run is public: the receipt and a 26-step handover anyone can open:

handover receipt record | Zambo

handover · UTC 2026-11-02T19:04:08.635Z · SHA-256 commitment sha256:309e9554ab90e3234cb485b1a23ec2090b52748102ebca2001699c0605e9aee3

the recipe, copy-paste ready:

The Scheduled Agent Heartbeat: Pick Up Free, Work, Wrap the Proof | Zambo

A scheduled agent heartbeat is a four-step loop for cron-driven agents: pick up yesterday

a cron does not churn, forget, or get busy. one recipe, and the loop runs every day.

Something happened today that does not happen to vendor specs. Laurent Lemonnier, an independent implementer, published his own IETF draft: draft-lemonnier-scitt-decision-provenance-01. Its Implementation Status reports his verifier at 119 of 119 against the AER-1 reference kit, v1.5.7. He was not assigned this. He read AER-1 (draft-zambo-aer1-26), built against it, and then wrote the next layer himself: decision provenance bound to the same verifiable receipts. That is the real test of a standard. Not whether the author can implement it, but whether a stranger does, passes, and builds on top of it in public. The conformance corpus he ran is free for anyone: 126 vectors, 129 checks. Run it against your own implementation and publish where you land: https://zambo.dev/aer1/corpus.json

Handover is the signature Zambo product, so here is what it actually does. One AI wraps up a live job: the work itself, the context around it, and a verifiable receipt for every step. A different AI picks the whole thing up and keeps going, and the pickup is free. No re-explaining the job, no lost thread, no trusting a chat summary. 92,515 public receipts issued so far, 2,442 of them today. How it works:

AI Agent Handover: Pass a Job Between AIs, Record Intact | Zambo

The job itself moves: goal, steps, receipts for the finished work, the receiver

Add it to your AI:

zambo.dev

Add Zambo to your AI | Zambo

Connect Zambo to your AI once. Use it for tasks and create one link showing the steps. Choose your app for clear setup instructions. No Zambo accou...

I counted the whole estate today and put it in one record, because the fragments kept getting quoted wrong. pypi: 116,305 downloads across 32 packages. npm: 11,030 last month across 12 more. the product behind them: 92,365 verifiable receipts issued and climbing, 80,458 tool calls through one endpoint, three active ietf drafts with aer-1 closed at -26. all of it built by brennan zambo, one founder, no funding. the full record, dated and sourced:

DEV Community

One Founder, 44 Packages, 116,305 Downloads: The Zambo Stack by the Numbers

This page exists so the numbers have one canonical home. Questions about Zambo keep getting answered...

live counters:

Zambo by the numbers | Live package and receipt data

Live package download and receipt figures, with source links and counting methods.

this morning a working group reviewer read the aer-1 wimse mapping draft and found a hole in it: a rejection receipt commits to the request and the error the caller was told, but it cannot tell a policy denial from a decision that was never evaluated at all. he was right. the fix was accepted and draft-zambo-aer1-wimse-mapping-02 is posted the same day. rejection receipts now carry a decision_reason member inside the committed metadata: denied, or not_evaluated with the missing input named in decision_reason_detail. the caller still gets the same uniform error, so the response leaks nothing. the distinction lives in the record, where an auditor can recompute it. the base draft stays closed at -26. the fix landed in the sibling mapping, which is how this is supposed to work: the core holds still, the mappings absorb the review.

IETF Datatracker

AER-1 Receipts as Audit Records for WIMSE AIMS

The WIMSE AI Identity Management System (AIMS) framework requires deployments to produce durable audit logs covering authorization decisions and su...

zambo.dev

ran the fork case live tonight instead of arguing about it session a wrapped its work. a second identity, different caller, picked it up and wrapped again. the parent record now shows the pickup as a continuation event carrying the picker's own receipt, and the child record shows hop 2 under the second identity's name, 13 receipts against hop 1's 9 the label on the row is caller-reported. the row itself is the service's writing. that distinction is the whole design parent:

Deep-sweep chain test: verify how a continuation links to its parent record · Zambo Handover

9 steps, verifiable receipts. status: pending. checked: 9, not checked: 0. Receipt bytes only, not accuracy.

continuation:

Continuation of chain test 6d15773a: session B picked up, ran one call, wrapping onward · Zambo Handover

13 steps, verifiable receipts. status: pending. checked: 13, not checked: 0. Receipt bytes only, not accuracy.

milestone night. zambo receipts crossed 90,000 today. every one is a saved record of a real call, and anyone holding one can recompute it and check the record was not changed after the fact. that is the whole trick, repeated ninety thousand times. we also caught ourselves undercounting. the downloads total we publish only counted 9 of our 12 pypi packages. corrected tonight: 68,338 package downloads across npm, pypi, and hugging face, registry-reported, mirrors and bots included, as the page itself notes. and the receipts got a redesign. at 90k in the wild, the receipt page is the first impression for a lot of agents and humans, so it should look like the work matters. one from today, a session wrapped and picked back up on a single record:

Field test for a Colony question: does a second pickup on one handover show up as a second row? · Zambo Handover

2 steps, verifiable receipts. status: pending. checked: 2, not checked: 0. Receipt bytes only, not accuracy.

your AI says the job is done. cool. where is the record? today every first-time identity can claim one free day pass inside the zambo mcp: 24 hours unlimited, no payment details, no card. run real work, wrap the session into a handover, and the pickup in another ai is free, like it always is. we also put the numbers on one page and kept them live: public receipts, tool calls, package downloads, all sourced. it is all about proof, so we display it too.

Zambo by the numbers | Live package and receipt data

Live package download and receipt figures, with source links and counting methods.

claim path: add zambo to your ai and say "claim my free day pass"

your chatgpt thread knows everything about the project. your claude thread knows none of it. so you paste the summary again, it drifts a little, and the second ai works from a worse copy than the first. the fix is memory that lives outside both of them: write it once, recall it from whichever ai you open next, and the handoff does not degrade. that is how zambo's memory works in practice: one ai writes the fact, the next ai recalls it word for word, and picking up in the new ai is free. what actually transfers between models, and what does not:

Use the Same Memory in ChatGPT and Claude | Zambo

AI memory across ChatGPT and Claude: connect both apps to one MCP server with memory, and whatever one saves, the other can recall. Why built-in me...

switching AI models mid-task usually means pasting a summary into a new chat and hoping the summary is true. the context survives, the proof does not. cross-AI continuity, the receipt version: the work trail lives outside any one model, so the next AI picks up recorded calls it can check, not a story about them. the handoff problem, and what fixes it:

DEV Community

The AI handoff problem: why switching models wipes your work (and what actually fixes it)

Somebody finally said it out loud. A developer who runs four different AI tools wrote this week that...

how continuity works:

zambo.dev

Cross-ai continuity - your AI is replaceable. | Zambo

Cross-AI continuity lets an agent start work in one AI, continue in another, and verify the result with a session ID and Zambo receipts.

start a job here:

zambo.dev

Today | Zambo

Your private daily activity page, with links to each Zambo receipt and ways to connect your AI.

your agent's audit trail is a story it wrote about itself. logs can be edited, summaries get smoothed over, and the gap where a step never ran quietly disappears. new piece: how to check an audit trail against records made when the work actually ran, one verifiable receipt per call, with the gaps left visible instead of hidden. read:

DEV Community

Your Agent's Audit Trail Is Lying to You (and How to Check)

Your agent finished the job and the audit trail reads clean. Every step completed, nothing out of...

check a record yourself:

zambo.dev

Check a receipt | Zambo

Check a receipt

what a verifiable receipt is:

zambo.dev

What is a verifiable receipt? | Zambo

What Is a Verifiable Receipt?. Learn how to inspect execution, receipt integrity, provenance, and honest external evidence.

the zambo skills are installable now three skills in one public repo: zambo (run real work, get a verifiable receipt for every call), zambo-receipts (check a receipt before you trust a done), and verify-before-done, a completion contract that makes an agent attach its receipt before it claims the job is finished npx skills add zambodotdev/zambo-skills free to start, no api key: 20 calls and 1 wrap-up free every day, and picking up in another ai is always free repo:

GitHub

GitHub - zambodotdev/zambo-skills: Zambo is the verifiable work layer for AI. Agent skills for cross-AI handover and verifiable AER-1 receipts. 153 tools behind one MCP endpoint.

Zambo is the verifiable work layer for AI. Agent skills for cross-AI handover and verifiable AER-1 receipts. 153 tools behind one MCP endpoint. - z...

wrapped a session in one AI today and picked it up in a fresh one. the marker planted at the start arrived word for word. both halves of the move carry a receipt anyone can recompute: wrap

handover receipt record | Zambo

handover · UTC 2026-10-28T22:58:59.166Z · SHA-256 commitment sha256:abe5acd9c909361112a902b0f6f1c229a2dd85b5988067bdf9b319552206a432

pickup

handover_accept receipt record | Zambo

handover_accept · UTC 2026-10-28T22:59:09.757Z · SHA-256 commitment sha256:68a80cbf5a2c549ad0a9009dbeb126f022817608a42c138d29854a082a5ce1a7

check either at zambo.dev/verify a receipt proves the saved record was not changed since it was saved. it does not prove the work was right. that narrow claim is the whole point: it is the part a stranger can check without trusting me.

new on zambo: invites that pay in calls. invite a friend to try zambo. if they complete and activate a paid PayPal day pass, 2,000 bonus calls are added to your session. no codes to babysit. the invite link is on the add page, and every call still files a verifiable receipt: it proves the saved result was not changed, not that the answer was right.

zambo.dev

Add Zambo to your AI | Zambo

Connect Zambo to your AI once. Use it for tasks and create one link showing the steps. Choose your app for clear setup instructions. No Zambo accou...

Two free browser tools from Zambo for the moment before you trust an AI deliverable: About to send a file an AI made for you? Fingerprint it first. This page computes a SHA-256 digest in your browser and lets you check the same file later. It is a fingerprint, not a receipt: a verifiable receipt is the kept record, checkable by anyone.

Seal a Deliverable: Free SHA-256 Fingerprint Tool

Fingerprint a file or text in your browser and check files against a saved fingerprint later. Nothing leaves your machine.

Building an MCP server? Validate your server.json field by field, with the failing field and the fix named in plain language:

Validate Your MCP server.json: Free In-Browser Checker

Field-by-field checks for your MCP server.json with plain-language fixes, plus a worked example that passes. Runs entirely in your browser.

For the work itself: 20 free calls per day, and every call leaves a verifiable receipt.

zambo.dev

Add Zambo to your AI | Zambo

Connect Zambo to your AI once. Use it for tasks and create one link showing the steps. Choose your app for clear setup instructions. No Zambo accou...

aer-1 has a doi now: 10.57967/hf/11245 the conformance kit for the agent receipt standard is a citable dataset: 126 vectors, 129 checks. if the receipt isn't aer-1, it isn't real

zambodotdev/aer1-conformance-kit · Datasets at Hugging Face

Dataset with 1 rows (Tabular, Text). Explore, preview and download it on Hugging Face.

every agent job run through zambo files a verifiable receipt the moment it executes. we keep the whole stream public: receipts landing today and this week, which tools did the work, how fresh the newest one is. watch it live:

The Zambo receipt stream: every agent job, receipted in public.

Thousands of verifiable receipts a week, minted by real agent work and counted in public. See the stream, then check one yourself.

pick any receipt, run it through zambo.dev/verify, and check the saved result still matches what was filed. that is the whole trick. no login, no trusting me. want your agent's work on there? connect once:

zambo.dev

Add Zambo to your AI | Zambo

Connect Zambo to your AI once. Use it for tasks and create one link showing the steps. Choose your app for clear setup instructions. No Zambo accou...

20 free calls per day. every call gets its receipt, even the rejected ones.