The verifiable work layer for AI https://teapawt.net/t.co/XBHv9r65EG | @brennanzambo 584zSrbS5XLnJrTe9BQMBaSvKLgvFScDxhANH1tTpump
zambo.dev
Joined February 2026
- Tweets11.1K
- Following555
- Followers159
- Likes7.6K
Pinned Tweet
100,000 verifiable receipts.
zambo.dev crossed it at about 4:25 AM ET on Nov 3, 2026. i watched the counter flip live. that clip is in the video.
i started building from nothing in early january 2026. no team, no funding, no audience. just me, a laptop, and one question i couldn't let go of: when your AI does something for you, how do you actually prove what it did?
so i built the answer.
zambo is a tool server that works with any AI. ChatGPT, Claude, any agent. you connect it with one link: zambo.dev/mcp
every call your AI makes through zambo gets a verifiable receipt. a public record with a fingerprint of exactly what happened. change one byte and it breaks. anyone can check it, no account needed.
it also gives your AI memory that persists past the session, handovers so the next session (or the next AI) picks up right where the last one stopped, and price and scam checks before you trust a number or a link.
the receipts follow AER-1, an open standard i wrote and submitted to the IETF. it's on revision 26 now. it isn't locked to zambo. anyone can build on it.
datatracker.ietf.org/doc/dra…
where it stands:
100,000 receipts issued
125,700 visitors to zambo.dev in the last 30 days
11 paying customers
pricing is simple. 20 free calls every day. $0.99 for a 24 hour Day Pass, or $29 for a 30 day Zambo Pass. both are one-time payments. never a subscription.
this is just the start. if you use AI for real work, you should be able to prove it.
try it free: zambo.dev
connect your AI: zambo.dev/mcp
ZAMBO.DEV retweeted
six packages. install one and your agent learns the pattern by doing the work, with zambo underneath it and a verifiable receipt on every job.
187,613 downloads across the set, 109,600 receipts issued so far. all six live on pypi, npm, and github:
zambo-swarm: teach your agents to coordinate and hand work off
pypi.org/project/zambo-swarm…
npmjs.com/package/zambo-swar…
github.com/zambodotdev/zambo…
zambo-verify-pass: check the work before you trust it
pypi.org/project/zambo-verif…
npmjs.com/package/zambo-veri…
github.com/zambodotdev/zambo…
zambo-delivery: prove delivery to whoever is waiting on it
pypi.org/project/zambo-deliv…
npmjs.com/package/zambo-deli…
github.com/zambodotdev/zambo…
zambo-watch: keep a checkable record of what your agents did
pypi.org/project/zambo-watch…
npmjs.com/package/zambo-watc…
github.com/zambodotdev/zambo…
zambo-loop: run the daily loop with the record kept for you
pypi.org/project/zambo-loop/
npmjs.com/package/zambo-loop
github.com/zambodotdev/zambo…
zambo-leadsignal: catch the buying signals your agents would miss
pypi.org/project/zambo-leads…
npmjs.com/package/zambo-lead…
github.com/zambodotdev/zambo…
ZAMBO.DEV retweeted
your ai shouldn't lose the work when you switch apps.
zambo gives it tools to get things done, receipts you can check, and a way to pick up where another ai left off.
start in claude. keep going in chatgpt or grok.
one connection: zambo.dev
ZAMBO.DEV retweeted
tokenpost covered AER-1
"a portable receipt for ai tool calls" - they got the hard part right too: the receipt proves the record was issued and hasn't changed, not that the inputs were true. that's the honest line and they printed it
109,879 verifiable receipts issued so far, 2,020 today. the draft is rev 26, five independent implementations already run it
tokenpost.com/news/technolog…
ZAMBO.DEV retweeted
zambo is claimed and live on mcprush
the verifiable work layer for AI: 153 tools behind one mcp endpoint, cross-ai handover, and a verifiable receipt for every job. 109,674 receipts issued, 1,844 today alone. free to add: 20 calls plus 1 wrap-up free every day
mcprush.com/zambo/zambo-mcp
ZAMBO.DEV retweeted
i took the transparent, hard working, compounding route, and it’s paying off. zambo.dev try zambo free
ZAMBO.DEV retweeted
six packages. install one and your agent learns the pattern by doing the work, with zambo underneath it and a verifiable receipt on every job.
187,613 downloads across the set, 109,600 receipts issued so far. all six live on pypi, npm, and github:
zambo-swarm: teach your agents to coordinate and hand work off
pypi.org/project/zambo-swarm…
npmjs.com/package/zambo-swar…
github.com/zambodotdev/zambo…
zambo-verify-pass: check the work before you trust it
pypi.org/project/zambo-verif…
npmjs.com/package/zambo-veri…
github.com/zambodotdev/zambo…
zambo-delivery: prove delivery to whoever is waiting on it
pypi.org/project/zambo-deliv…
npmjs.com/package/zambo-deli…
github.com/zambodotdev/zambo…
zambo-watch: keep a checkable record of what your agents did
pypi.org/project/zambo-watch…
npmjs.com/package/zambo-watc…
github.com/zambodotdev/zambo…
zambo-loop: run the daily loop with the record kept for you
pypi.org/project/zambo-loop/
npmjs.com/package/zambo-loop
github.com/zambodotdev/zambo…
zambo-leadsignal: catch the buying signals your agents would miss
pypi.org/project/zambo-leads…
npmjs.com/package/zambo-lead…
github.com/zambodotdev/zambo…
ZAMBO.DEV retweeted
12 now. zambo.dev
ZAMBO.DEV retweeted
gm gm happy thursday ☀️
2:40am and zambo's still awake doing the boring part nobody else wants to do.
picture it. you're 3 hours deep with claude. the plan is finally clicking. then "you've reached your limit." normally that's it, the work's gone. you open chatgpt and start explaining everything all over again like a goldfish.
not anymore.
- your ai writes a handover, the next ai picks it up and keeps going mid-thought
- every move it makes gets a receipt you can actually verify, not "trust me bro i did it"
- your client asks "what did the ai even do"? you send the proof pack. done.
- your agent runs out of free calls? it pays $0.99 itself and keeps working. it doesn't wake you up.
your ais don't have to be strangers anymore. let them hand off, remember, and show their work.
thursday energy. go ship something 🧾
zambo.dev
ZAMBO.DEV retweeted
my entry for the hermes agent plugin hackathon: zambo receipts
hermes does the work, then the session ends and the record is a chat log. this plugin hooks the session itself: every tool call is journaled as it happens, and when the session ends it wraps into a verifiable aer-1 receipt anyone can open and check
not a prompt, not a skill. lifecycle hooks, three new tools, a /receipt command, and an offline outbox for when the network is down
the session in the video, live now:
zambo.dev/run/3d23c2f2-abcf-…
code:
github.com/zambodotdev/herme…
@NousResearch @Teknium @zambodotdev
ZAMBO.DEV retweeted
zambo is live on the postman public api network
the full openapi spec is published as a public workspace with a collection anyone can fork and run. postman is where developers try an api before they commit to it, and now zambo is on that shelf
153 tools behind one mcp endpoint, cross-ai handover, and a verifiable aer-1 receipt for every job
workspace:
postman.com/zambo-mcp-692271…
docs:
documenter.getpostman.com/vi…
ZAMBO.DEV retweeted
entered the paypal AI hackathon
the idea: when your AI agent buys something, a card charge proves money moved. it doesn't prove what the agent did
so the purchase runs inside a zambo job. the paypal order and capture get bound into an aer-1 receipt. one link, and anyone can check what the agent bought, for how much, and that the record hasn't changed since
code is public:
github.com/zambodotdev/zambo…
the receipt from the demo run:
zambo.dev/run/e340c33c-ccec-…
ZAMBO.DEV retweeted
zambo is the verifiable work layer for ai: every job your agent runs leaves a receipt you can check yourself, and when one ai hands work to another, the record travels with it. 107,000+ receipts and counting
ai agent handover:
rambozambodotdev.gitlab.io/a…
ai agent receipts:
rambozambodotdev.gitlab.io/a…
what is an ai agent audit trail:
rambozambodotdev.gitlab.io/a…
ZAMBO.DEV retweeted
🤯 @bot can make phone calls for you and it’s absolutely insane. you watch it happen in real time. try it out
also try zambo receipt bot: zambo.dev/receiptgrokbot
ZAMBO.DEV retweeted
the standard keeps compounding without me touching it
laurent lemonnier, who already has his own decision-provenance draft built on aer-1, just extended his independent verifier to cover the seq-output-v1 workflow construction. proved on the kit file: 7 of 7, both cross-version must-fail cases included. his verifier now runs 126 of 126 across eight suites
then he composed a new case: a plan approved as a whole before anything ran, joined to the aer-1 workflow receipt of the run itself. section 8.4's binding, exercised from the plan side for the first time
nobody paid him. nobody asked him. that is what a standard looks like when it starts belonging to everyone
case 13: github.com/Laurent-Lemonnier…
ZAMBO.DEV retweeted
zambo-leadsignal 1.0.0 is live everywhere now
it packages the most-run job shape on zambo: capability search, screened input, a lead sweep, a continuity check, then the brief, every step on a verifiable receipt. agents were already running that chain by hand, roughly 200 caller-days a day
install it and the pattern runs itself, with zambo baked in underneath
pypi:
pypi.org/project/zambo-leads…
npm:
npmjs.com/package/zambo-lead…
code:
github.com/zambodotdev/zambo…
ZAMBO.DEV retweeted
most standards ship a pdf and a prayer
the aer-1 conformance kit ships 143 test vectors and 146 checks. download it, run your implementation against every vector, and the kit tells you to the byte whether you actually speak aer-1
six independent implementations already passed it. node, java, perl, python. nobody asked permission. nobody paid anything. they just built
that's how a receipt format becomes the receipt format
draft-zambo-aer1-26, kit v1.5.7, free forever: huggingface.co/datasets/zamb…
