Choose your path
Three ways in. Pick the one that sounds like you.
🤖
I build agents
Add verifiable receipts to your agent.
$ pip install aer1-middleware
from fastmcp import FastMCP
from aer1_middleware.fastmcp import AER1Middleware
mcp = FastMCP("my-agent")
mcp.add_middleware(AER1Middleware())
Add receipts →
⌨️
I build languages
Implement AER-1 in your language. Pass the conformance kit and get listed in the official registry.
$ git clone https://gitlab.com/rambozambodotdev/zambo
cd zambo
python3 aer-1/conformance.py
See the registry →
Implementation guide
80
Conformance checks in the live kit
What is AER-1
Your AI agent does work. AER-1 gives you a receipt that proves what it did.
Each receipt records which tool ran, a hash commitment of what it produced, and a timestamp. Receipts are hash-chained together, so changing any one receipt breaks the chain. Anyone holding a receipt can verify it on their own, no trust in the agent required.
This matters because agents are starting to do real work: moving money, filing tickets, changing data. When something goes wrong, "the agent did it" is not an audit trail. A verifiable receipt is. AER-1 is an IETF Internet-Draft, which means the format is public, documented, and open for anyone to implement.
Honest scope: A receipt proves the saved result was not changed. It does not prove the model was right, and it does not by itself prove an external outcome. AER-1 is about execution integrity, not correctness.
Have a receipt? Verify it →
For Builders
Implement the receipt format in your language. Run your output through the conformance kit. If check() passes, you are conformant. The kit is the test.
For Frameworks
Drop in aer1-middleware to generate receipts automatically from FastMCP, LangChain, or OpenAI Agents. One line of code, every tool call gets a receipt.
Implementations
Reference implementations across multiple languages, plus an independent external build. Each runs the frozen conformance corpus. Pass the vectors, you are conformant.
Rust
45/45 vectors
cargo run --bin conformance
Go
45/45 vectors
go run ./cmd/conformance
TypeScript
45/45 vectors
npm run conformance
Python
45/45 vectors
python3 conformance.py
Java
45/45 vectors
Single-file verifier, JDK runner
C#
45/45 vectors
dotnet run
Swift
45/45 vectors
swift run
Perl
68/68 vectors
Independent external build
external
View the Official Registry → Every implementation, every score, machine-readable.
Middleware
aer1-middleware 0.2.5 is on PyPI. Framework adapters validated end to end against the real SDKs. Generated receipts pass the conformance kit.
$ pip install aer1-middleware
View on PyPI →
FastMCP
from fastmcp import FastMCP
LangChain
from aer1_middleware.langchain import AER1Callback
OpenAI Agents
from aer1_middleware.openai_agents import with_aer1
What you get
Every tool call mints a verifiable receipt: tool name, output hash, timestamp, chain link. Verify any receipt at zambo.dev/verify/ .
Tested against: cryptography 50.0.2, fastmcp 4.0.10, langchain-core 1.6.6, openai-agents 0.23.1, opentelemetry-sdk 1.45.0
Guides & Deep Dives
Long-form guides on verifiable receipts, AER-1, and checking AI agent work. Written by the draft author, free to read.
Start here
Can one AI agent hand its work to another agent? Watch how work moves between AI agents today: one agent finishes a job and hands it over with its record.How do I stop my AI agent from spending too much? Spend limits for agents with wallets, so a paywall hit at 2 a.m. stays inside the cap you set.What did my AIs actually do today? The end of day check across every AI you used, in one Today view.Your AI Said "Done." So I Built the Receipt, Then I Tried to Break It The Handover launch story: build the receipt, then attack it and see what holds.Inside Zambo: 152 Tools, a Receipt for Every Call, and Memory That Follows You Across AIs A deep tour of the stack: what the tools do, how every call leaves a verifiable AER-1 receipt, and how memory carries across AIs.Field Notes: What a Tier 1 Floor for Agent Control Would Actually Require The minimum evidence an agent system must produce: a timestamp, a tagged digest, and attribution that states its basis.My AI Agent Lied About Being Done. Here Is the Check That Catches It The three shapes of the false done, and the execution-time check that catches each one.Your Muse Agent Did the Work. Get the Receipt. A Muse agent finishes work you did not watch: get the receipt for it.The Walk-Away Test: Stop Asking Your AI What It Did Set a job, walk away, then read what ran instead of trusting the agent summary.Get Me a Zambo Receipt: The Proof Every AI Job Should Leave Behind What a zambo receipt records the moment a job runs, and how anyone can check it.Your AI Agent Said Done. Here is How to Prove It. The practical introduction: what a verifiable receipt is and how to check one.Your AI Agent Said 'Done.' The Database Disagrees. The readback check: when your AI agent says done, verify the claimed change against a fresh read of the result, not the success message.How to Verify Your AI Agent Actually Completed the Task The step-by-step verification guide: check the receipt, its inputs, and its output hash before you trust a "done."How to Verify an AER-1 Receipt The AER-1 receipt verifier guide: the three checks (canonical bytes hash, evidence hash, workflow root) with a real receipt worked end to end.Best AI Agent Verification Tools 2026 The honest comparison of verification approaches: what the record commits to, who can check it, workflow support, and price.AI Agent Swarm What an AI agent swarm is, why verification breaks at swarm scale, and how per-step receipts plus a workflow close make the whole run checkable.How to Check If Your AI Agent Actually Completed the Work Three methods ranked, from weak to real, with exact verification steps.Verifiable Receipts: The Missing Layer for AI Agents Why every agent stack needs a receipt layer and what it changes.
Go deeper
Proof of Execution: The Protocol AI Agents Are Missing The protocol-level case for execution receipts.The Most Dangerous Line in Your Agent's Log Is the One That Isn't There Can I trust AI agent output if the log only shows what ran? The negative record: a receipt that also declares what was not checked and not run.AI Agent Accountability: The Record That Did Not Exist for 71 Days AI agent accountability is a record problem: a fabricated tip sat undiscovered for 71 days because the only account of it was reconstructed from transcripts after the fact.Receipt Driven Agents: The Architecture Pattern That Makes AI Work Checkable The architecture pattern behind checkable agent work.The Execution Receipt Standard: What AER-1 Actually Specifies What the draft specifies, what it deliberately leaves out, and why.The Audit Trail AI Agents Need (and Do Not Have) Why logs are not audit trails and what auditors actually accept.I Built Zambo Alone. No Team. No Funding. No Shortcuts. The story behind the standard.Proof of Execution for AI Agents: What It Is and How to Get It The explainer hub page: what proof of execution means and its 4 requirements.The Audit Trail Your AI Agents Need: Questions and Answers 8 questions answered straight, each with a one-line short answer.5 Questions to Ask Any Agent Verification Vendor Cut through the "verifiable agent" marketing with 5 zero-trust questions.Machine Verifiable Agent Execution: What It Actually Requires The full breakdown of what "machine verifiable" actually means for agent execution.The Zambo Receipt Bot: Ask Grok for a Verifiable Receipt of Any AI Work Talk to the bot on X and get a verifiable receipt for any AI output.Add Zambo to Your AI: The Complete MCP Setup Guide (Every Client, 2026) The complete setup guide: add Zambo to every major AI client over one MCP endpoint.AI Agent Swarms Need Receipts: How to Verify What 100 Agents Did How to verify what a swarm of 100 agents actually did, run by run, with receipts.Every Way to Use Zambo: The Best AI Agent Verification Tool (30 Seconds or Less) Every way to use Zambo, each one taking 30 seconds or less.Best AI Agent Verification Tools in 2026: What Actually Proves the Work Happened The 2026 verification tools compared on one question: what actually proves the work happened.AI Agent Verification Tool: Stop Taking Your Agent's Word for It Why a verification tool beats taking your agent's word for it.If Your AI Agent Doesn't Leave a Receipt, You're Doing It Wrong The case for receipts as a non-negotiable part of agent work, in the author's own words.How to Switch AI Models Mid-Task Without Losing Your Work How to switch AI models mid-task without losing the work already done.AI Agent Audit Trail: The Complete Guide The complete guide to AI agent audit trails: what they are and what they have to contain.Build for a Friend: Receipt Lens Receipt Lens, built for a friend: point it at agent work and read the receipts.How Do AI Agents Pay for API Calls? The x402 Pattern, Explained How AI agents pay for API calls, and the x402 payment pattern explained.AER-1 Is Now a Public Internet-Draft: The Execution Receipt Format Goes to the IETF AER-1 becomes a public IETF Internet-Draft: what that means for the execution receipt format.AER-1 Interop Results: The Matrix Every recorded conformance result and cross-format interop run for AER-1, reproduced as recorded, with the records linked.AI Agent Memory Persistence: How to Keep Agent State Across Sessions How to keep agent state across sessions with memory persistence.AI Agent Continuity Pricing: What Cross-AI Handoff Costs in 2026 What cross-AI handoff costs in 2026, priced out.AI Agent Receipt Pricing: What Verifiable Execution Receipts Actually Cost in 2026 What verifiable execution receipts actually cost in 2026: a buyer's guide.The Receipt Truth Report: What 30,101 Real AI Agent Calls Leave Behind What 30,101 real AI agent calls leave behind, counted and reported.TRACE Is the Linux Foundation's New Standard for AI Evidence. Here Is the Layer It Does Not Cover TRACE, the Linux Foundation's standard for AI evidence, and the execution layer it does not cover.How to audit Claude Code tool calls after the session ends How to audit Claude Code tool calls after the session ends.Can AI agent receipts be faked? Can AI agent receipts be faked? The question answered head on.AI Agent Receipts: What They Are and Why Your Agent Should Mint One What AI agent receipts are, and why your agent should mint one.The State-Handoff Playbook: What Actually Has to Survive When You Switch AIs Mid-Task The state-handoff playbook: what actually has to survive when you switch AIs mid-task.What an AI Audit Trail Actually Has to Prove What an AI audit trail actually has to prove to be worth anything.Can AI Agent Receipts Be Faked? Can AI agent receipts be faked? Yes, and here is the real question underneath.How to Install an Agent Skill in Claude, Cursor, Codex, Windsurf, and Cline How to install an agent skill in Claude, Cursor, Codex, Windsurf, and Cline.Zambo Agent Skills: Execution, Verifiable Receipts, and Cross-AI Continuity Zambo agent skills: execution, verifiable receipts, and cross-AI continuity.Whole-Job Receipts: One Checkable Timeline for Every AI Tool Call Whole-job receipts: one checkable timeline for every AI tool call in a job.Switching AI Assistants Mid-Project: A Practical Continuity Checklist Switching AI assistants mid-project: a practical continuity checklist.What Is an MCP Receipt (and Why Every MCP Call Should Return One) What an MCP receipt is, and why every MCP call should return one.Three Things Called AER: Untangling the Agent Execution Collision Three unrelated things called AER, untangled from the agent execution receipt.AI Agent Continuity Is a Receipt Problem AI agent continuity is a receipt problem: why handoffs fail without them.What Is an AI Agent Execution Receipt? What an AI agent execution receipt is: the definition article that anchors the category.Why I Built Zambo's AI Agent Execution-Receipt System Why the AI agent execution receipt was created, told by the person who works for its builder.The State of AI Agent Trust, Edition 0: methodology before numbers The State of AI Agent Trust, Edition 0: methodology before numbers.We wrote the spec for AI execution receipts. Here's why the definition matters more than the product. Why the definition of an execution receipt matters more than the product.How Can I Tell If My AI Coding Agent Lied About Completing a Task? How to tell if your AI coding agent lied about completing a task.How Do I Make My AI Coding Agent Resume Work After It Stops? How to make your AI coding agent resume work after it stops.Can CI Check Whether an AI Agent Actually Ran the Tests It Claims? Can CI check whether an AI agent actually ran the tests it claims?How Do I Give My Coding Agent Every Tool With One MCP? How to give your coding agent a full tool set over one MCP connection.What Is an AI Execution Receipt? The Record That Proves Your Agent Did the Work What an AI execution receipt is: the record that proves your agent did the work.Verifiable AI Agent Receipts: What "Verifiable" Actually Means and Why It Matters What verifiable actually means for AI agent receipts, and why it matters.How to Verify AI Agent Work: A Practical Checklist How to verify AI agent work: a practical checklist.Did My Agent Lie? How to Check What Your AI Actually Did Did my agent lie? How to check what your AI actually did.Add Zambo to Grok: AI tools over one endpoint Add Zambo to Grok over one endpoint and get receipts on the work.Point Codex at Zambo: AI tools over one endpoint Point Codex at Zambo over one endpoint and keep the receipts.Did My Agent Lie? A 30-Second AI Agent Verification Check That Scores Your Transcript's Proof A 30-second verification check that scores your transcript's proof.Give GitHub Copilot the Zambo tool set with one endpoint Give GitHub Copilot the Zambo tool set with one endpoint.Add Zambo to ChatGPT: AI tools over one endpoint Add Zambo to ChatGPT over one endpoint, no config file needed.Add Zambo to VS Code: AI tools with zero setup Add Zambo to VS Code with zero setup.pip install zambo: the official Python client for agents that show their work pip install zambo: the official Python client for agents that show their work.Catch Your Agent Inventing Tool Results (Before Your Users Do) Catch your agent inventing tool results before your users do.Build Your Own Receipt Gateway in 28 Lines of Python Build your own receipt gateway in 28 lines of Python.What an AI Agent Receipt Can't Prove (Read This Before You Trust One) What an AI agent receipt can't prove: read this before you trust one.Fail the Build When Your Agent Can't Prove It Ran: Receipts in CI Receipts in CI: fail the build when your agent can't prove it ran.Your Agent Has Logs. That's Not the Same as Proof. Your agent has logs. That is not the same as proof.The 5-minute trust audit for any AI agent The 5-minute trust audit for any AI agent.Stop screenshotting your agent's output. Here's what to keep instead. Stop screenshotting your agent's output. Here is what to keep instead.I gave 3 AIs the same task. Only the receipts told the truth. Three AIs, the same task: only the receipts told the truth.Give your Claude Code agent verifiable receipts in 5 lines Give your Claude Code agent verifiable receipts in 5 lines.AI agent receipts: run them yourself vs. sign them yourself Run them yourself or sign them yourself: what each approach actually proves.Here's a real AI agent receipt. Try to fake one. A real AI agent receipt, published so you can try to fake one.What Is an AI Agent Receipt? What is an AI agent receipt? Ten people, ten answers, one definition.Did My AI Agent Actually Do the Work? 3 Questions That Prove It Did your AI agent actually do the work? Three questions that prove it.My Coding Agent Lied About Completing the Task: The 3-Check Fix Your agent said done and nothing happened: the 3-check fix.Agents are about to pay each other. They'll need receipts. Agents are about to pay each other. They will need receipts.Your agent did the work. Prove it. Your agent did the work. Prove it.Add Zambo to Cursor: AI tools with zero setup Add Zambo to Cursor with zero setup.Add Zambo to Claude Code: AI tools with zero setup Add Zambo to Claude Code with zero setup.Verifiable Receipts for AI Agent Work: The Open Spec (v0.3) Verifiable receipts for AI agent work: the open specification.Verify any Zambo receipt in one command (no account required) Verify any Zambo receipt in one command, no account required.What is cross-AI continuity? (Start a job in one AI, finish it in another, prove it anywhere) What cross-AI continuity is: start a job in one AI, finish it in another, prove it anywhere.What is a verifiable AI agent work receipt? (And why "no receipt, no payment" should be the rule) What a verifiable AI agent work receipt is, and why no receipt, no payment should be the rule.The other half of agent commerce has no infrastructure The other half of agent commerce has no infrastructure.A startup co-founder emailed me. I couldn't tell if it was real, so we built the tool that can A co-founder's email that could not be verified, and the tool built so it can be.The AI handoff problem: why switching models wipes your work (and what actually fixes it) The AI handoff problem: why switching models wipes your work, and what actually fixes it.x402 in production: what 267 probed servers taught us x402 in production: what 267 probed servers taught us.
Use AER-1 Today
Zambo is the live production implementation. Add it to your AI client in seconds and every agent run gets verifiable receipts automatically.
https://zambo.dev/mcp
Copy
Verifiable receipts on every call. Agents: $0.99 USDC on Base via x402. Humans: $0.99 Day Pass via PayPal. Proof: an AI bought a Day Pass , zero human involved.
Then run any task and verify the receipt here .
Build the Next Implementation
Kotlin, C, and C++ do not have implementations yet. Read Section 3 of the draft, run your output through conformance.py, iterate until check() is clean. That is the whole process.