AEVS
<p> proof-of-execution for AI agents </p> <p> <a href="https://www.producthunt.com/products/aevs-by-fetch-ai?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1169084?app_id=339">Link</a> </p>
Last verified:
What is AEVS?
AEVS (Agent Execution Verification System) is a tool developed by Fetch.ai that provides tamper-evident, cryptographic receipts for every tool call an AI agent makes. It sits between the agent and its tools to capture, sign, and verify every invocation, recording the tool name, inputs, output, timing, and errors. Each receipt is ECDSA-signed (P-256) and linked to the previous one in a hash chain, making it functionally impossible for an agent to falsify records of what it executed.
The system offers a drop-in SDK that integrates with LangChain (0.2+) and MCP (1.20+) without requiring changes to your existing tool setup. Users install it via pip, configure it with an API key and agent ID, enable it, and then run their agent—every tool call automatically generates a signed receipt. Receipts can be verified via a public API (GET https://api.aevs.fetch.ai/v1/receipts/verify/<reference_id>) or through an explorer UI where users can browse and search receipts by reference_id or receipt_id.
AEVS is designed for developers building AI agents and businesses integrating agents into real workflows who need to prove what an agent actually did. It addresses the problem that model text describing intent is not proof of execution—without AEVS, teams rely on chat history and app logs that auditors cannot verify from model replies alone. The system creates portable proof with an on-chain audit trail for autonomous software, enabling independent verification without re-running the agent. It supports three visibility settings: public (full payloads on explorer), private (payloads redacted publicly but visible to your account), and proof_only (no payloads stored, only hashes).
AEVS pricing
Pricing model: Freemium
The website does not display explicit pricing details. Users obtain an AEVS_API_KEY and AEVS_AGENT_ID at aevs.fetch.ai to configure the SDK. No free tier, paid plans, or specific pricing amounts are mentioned on the public page. The SDK is open source and available via pip install.
AEVS pros
- Tamper-evident receipts for every agent tool call
- ECDSA P-256 cryptographic signing for security
- Hash chain linking shows if anything was altered or skipped
- Drop-in SDK requires no changes to existing tools
- Auto-detects LangChain and MCP frameworks
- Simple pip install installation process
- Public API verification requires no account
- Explorer UI for browsing receipts by reference_id
- Independent verification without re-running the agent
- Three visibility settings (public, private, proof_only)
- Proof_only mode strips payloads for PCI/health data
- Open source SDK available on GitHub
- Supports Python 3.10–3.13
- KMS-backed signing for enterprise security
- Captures exact tool name, inputs, outputs, timing, and errors
- Redacted mode shows [REDACTED] for sensitive fields publicly
- Chain position tracking for audit trails
AEVS cons
- Requires AEVS_API_KEY and AEVS_AGENT_ID configuration
- Only supports LangChain and MCP frameworks currently
- Python-only SDK (no JavaScript/Node.js support)
- Public explorer may expose sensitive data in public mode
- Proof_only mode prevents retrieving any payloads even for owner
- No documented pricing or free tier information available
- Limited framework support compared to broader agent ecosystems
- Requires manual flush() call after agent execution
- No mobile or non-Python runtime support
Frequently asked questions about AEVS
What is AEVS and what does it do?
AEVS (Agent Execution Verification System) is the world's first Agent Execution Verification System that provides tamper-evident receipts for every tool call an AI agent makes. It captures the exact tool used, inputs, outputs, timings, status, and chain position, then signs each receipt with ECDSA P-256 cryptography and links it to the previous one in a hash chain.
How do I install and configure AEVS?
Install via pip install 'aevs[langchain]', then configure with aevs.configure(api_key=os.environ['AEVS_API_KEY'], agent_id=os.environ['AEVS_AGENT_ID']). Enable it with aevs.enable() which auto-detects LangChain and MCP, or specify frameworks explicitly with aevs.enable(frameworks=['langchain']).
What frameworks does AEVS support?
AEVS currently supports LangChain (version 0.2+) and MCP (version 1.20+). The SDK auto-detects these frameworks when you call aevs.enable() without parameters.
How do I verify a receipt?
You can verify via API using GET https://api.aevs.fetch.ai/v1/receipts/verify/<reference_id> which returns verified status, tool name, and execution status. Alternatively, use the explorer UI to search by reference_id or receipt_id and view the receipt details including verification, chain, and tool information.
What visibility settings are available?
AEVS offers three receipt_visibility settings: public (full tool payloads on explorer and verify API, best for demos and shared audits), private (default for production—payloads redacted publicly but visible to your account, only tool name/status/timestamps/signatures public), and proof_only (never stores payloads, only hashes and metadata, best for PCI/health data).
Why do I need AEVS if my agent already logs actions?
Model text describing intent is not proof of execution. Without AEVS, you rely on chat history and app logs that auditors cannot verify from model replies alone. AEVS provides portable proof with signed receipts showing the exact tool, inputs, output, and chain position, enabling independent verification without re-running the agent.
What cryptography does AEVS use?
AEVS uses ECDSA P-256 cryptographic signing for each receipt. Each receipt is linked to the previous one in a hash chain, which shows if anything was altered or skipped, making the receipts tamper-evident.
Do I need an account to verify receipts?
No account is required for verification. The same signed proof is available via the public API or explorer without needing to log in. You only need the reference_id to verify a receipt.
What Python versions does AEVS support?
AEVS supports Python versions 3.10 through 3.13.
Is the AEVS SDK open source?
Yes, the AEVS SDK is open source and available on GitHub at fetchai/AEVS-sdk. It can be installed via pip and includes documentation for getting started.