Zambo Receipt Bot: Using Cryptography to Prove AI Actually Did the Work

AI Says "Done" — But Did It Actually Do It?

An AI agent tells you a task is complete. But what exactly did it do? Was the output tampered with? When did it finish? As AI takes on more and more real-world work, these questions are becoming an unavoidable trust problem.

Zambo's answer is a "Receipt Bot" running on X (formerly Twitter). The core logic is straightforward: when an AI completes a task, instead of giving you a verbal promise or a screenshot, it generates a verifiable cryptographic receipt. This receipt proves three things — that the work actually happened, that the output hasn't been altered since, and the exact time the task was completed.

What the Receipt Bot Actually Does

The bot runs directly on X. Ask it to handle any AI task, and it generates a verifiable receipt. The developers emphasize this isn't a screenshot or a promise — it's a cryptographic credential that anyone can independently verify.

The key is the open specification behind it. Every receipt follows the AER-1 open standard, an Internet-Draft submitted by the authors to the IETF, which defines a portable format for AI execution receipts. Any third-party verifier can check the authenticity of any receipt through Zambo's verification page at zambo.dev/verify.

This open-standard design is the foundation of the entire approach. It means receipts aren't locked into a closed platform but follow a publicly auditable technical specification — in theory, any tool implementing the standard can generate and verify receipts.

Background: AER-1 (AI Execution Receipt) is a receipt format specification designed specifically for AI task execution. Submitting an Internet-Draft to the IETF signals the authors' intent to pursue it as an internet standard, though it currently remains in draft form and has not yet become an official RFC. The IETF (Internet Engineering Task Force) is the international standards body behind core internet protocols like TCP/IP, TLS, and HTTP. Its standardization process is open and transparent — anyone can submit drafts and participate in discussions. Getting a specification accepted as an RFC typically involves working group review, broad public comment periods, and final review stages, often spanning several years.

The core technical principle behind cryptographic receipts works like this: the moment a task is completed, the output is hashed (creating a digital fingerprint of the content), timestamped, and then signed with a private key — all packaged together as a receipt. Anyone can verify the signature using the corresponding public key, then recompute the hash of the current content and compare it against the hash recorded in the receipt to determine whether the content was tampered with and whether the signature is authentic. This mechanism is the same integrity-verification principle used in HTTPS certificates and blockchain transactions — a well-established cryptographic technique.

Four Real-World Use Cases

The developers outline several representative scenarios, all pointing to the same pain point: how to establish verifiable trust in AI collaboration.

Freelance Delivery Proof

Suppose you hire an AI agent to write 50 product descriptions and it claims the job is done. Instead of just trusting it, you can use the Receipt Bot to generate a verifiable receipt for each description. Clients can check them one by one on the verification page at any time, eliminating delivery disputes at the source.

Agent-to-Agent Task Handoffs

When a research Agent completes its work and hands off to a writing Agent, the receiving agent shouldn't blindly trust the output. It can check the receipt — making the handoff "verified" rather than "assumed." This is especially important in multi-agent collaborative systems and serves as a foundational guarantee for reliable AI workflows.

Background: Multi-Agent Systems are a major direction in current AI development. In these architectures, specialized agents handle research, planning, writing, review, and other steps in sequence, forming a pipeline. This design dramatically improves handling of complex tasks, but introduces a new risk: each agent can only "see" what the previous agent passed to it, with no way to directly verify whether that output accurately reflects the upstream work. Without verifiable receipts, an error or "hallucination" from one agent can be directly adopted and amplified by downstream agents, creating an error chain. This is why task handoff verification is considered a foundational requirement — not an optional feature — in multi-agent scenarios.

Compliance and Audit Trails

In regulated industries, every AI decision needs a traceable record. The Receipt Bot provides timestamped, tamper-evident credentials that auditors can independently verify — no need to rely on a company's internal logs to "prove their own innocence."

Dispute Resolution

When a client claims the AI didn't complete the work, pull up the receipt. Cryptographic evidence settles the case, eliminating endless back-and-forth "he said, she said" arguments.

Why This Matters

AI handles more work every day, yet there's currently no standardized way to prove "what actually happened." Screenshots can be faked, logs can be edited, and "trust me" clearly isn't enough.

What makes verifiable receipts different is when and how they're generated: at the moment of execution, cryptographically sealed, independently verifiable by anyone. This transforms "trust" from a social convention into a fact that can be verified through technical means. As AI agents increasingly perform consequential actions on behalf of humans, this kind of verifiability is shifting from "nice to have" to a hard requirement.

How to Use It and Product Positioning

The barrier to entry is kept low: go to the bot's page on X, start a conversation, and ask it to verify an AI task or generate a receipt. No registration, no credit card required. The bot runs on Zambo's infrastructure and offers 20 free calls per tool per day. Developers who want to integrate receipt capabilities directly into their own applications can use the MCP interface.

The Receipt Bot is just one part of Zambo. The company positions itself as a "trust layer for AI work" — every AI task leaves a publicly verifiable receipt, and work records can follow users across different AI platforms. A $0.99 Day Pass is also available for light users who need occasional access.

From an industry perspective, this type of product reflects an emerging trend: as AI evolves from an assistive tool into an autonomous executor, the audit and accountability infrastructure around its behavior is becoming a new competitive frontier. If efforts like AER-1 — which is pursuing the IETF standardization route — gain broader adoption, it could become a common language for AI work credibility. Of course, whether an open standard truly gets adopted by the ecosystem still depends on uptake and multi-party implementation. For now, this remains an early-stage effort driven by a single team.

Background: MCP (Model Context Protocol) is an open protocol proposed by Anthropic in late 2024, designed to establish a unified interface standard for interactions between AI models and external tools and data sources. Developers only need to implement a server following the MCP spec to allow MCP-compatible AI clients (such as Claude) to call it directly — no need to build separate adapters for each AI platform. Zambo's MCP interface means developers can embed receipt generation capabilities directly into their AI workflows or agent applications, triggering receipt creation at the code level rather than through X's bot chat interface. This is especially useful for scenarios that require batch receipt generation or deep integration of verifiability into a product.