App and agent security · by Fetch.ai
AEVS (Agent Execution Verification System)
A Python SDK plus hosted backend that records every tool call a LangChain/LangGraph or MCP agent makes as a signed, hash-chained receipt, which anyone with the receipt ID can check through a public API or explorer.
A model saying it refunded a customer is not proof; AEVS gives you a checkable record of which tool ran, with what inputs and output, without changing your tools.
APIOpen sourceLicense unclearFor your project
- Official API
- Official API Beta; APIs and explorer may change. SDK truncates receipt payloads above 1 MB by default
“Returns receipt metadata, hash-chain status, and KMS signature info.” aevs.fetch.ai/llms.txt
- Key
- Needed
“Credentials from the dashboard at https://aevs.fetch.ai (sign up, register an agent)” aevs.fetch.ai/llms.txt
- Formats
- JSON
- Data download
- Not stated
- Price
- No free tier stated
- Open source
- fetchai/AEVS-sdk
- License
- Apache License 2.0 (SDK); no service terms page found The Python SDK is Apache 2.0; terms for the hosted receipt backend and dashboard are not published on the site
- Commercial use
- Commercial use unclear
- Credit
- Not stated
- Embedding
- Embedding unclear
- Using the website
- Free
“Verify via API or explorer No account required. Same signed proof either way.” aevs.fetch.ai/
Read on AEVS (Agent Execution Verification System)'s own pages on Oct 24, 2026 · license or terms · API docs · source code. Check the license page before you ship.
For your agent
A request your code or agent can make (add your key):
https://api.aevs.fetch.ai/v1/public/receipts/stats- pip install "aevs[langchain]" or "aevs[mcp]" (Python 3.10+), set AEVS_API_KEY and AEVS_AGENT_ID from the dashboard, call aevs.configure() then aevs.enable(), and aevs.flush() before exit; verify a receipt with GET https://api.aevs.fetch.ai/v1/receipts/verify/<reference_id>, no auth.
- Use receipt_visibility "private" or "proof_only" for personal or regulated data; "public" exposes full inputs and outputs to anyone with the reference_id
- Call aevs.flush() before a short-lived script exits or recent receipts may not be sent
- A receipt proves the SDK recorded the call, not that the external system performed it
llms.txt: aevs.fetch.ai/llms.txt
Paste into your agent: Read https://agentgid.com/resources/aevs/agent.md before using AEVS (Agent Execution Verification System).
Quick answers
Does AEVS (Agent Execution Verification System) have an API?
Yes. AEVS (Agent Execution Verification System) has an official API documented at https://github.com/fetchai/AEVS-sdk/tree/main/docs.
Does the AEVS (Agent Execution Verification System) API need a key?
Yes, you need an API key. Limits: Beta; APIs and explorer may change. SDK truncates receipt payloads above 1 MB by default
What you can build
- A payments or refunds agent that shows the user a receipt link for every action it took
- An audit page for an internal agent that lists signed tool-call receipts per run, using proof_only mode for sensitive data
- A CI check that replays an agent task and fails if any receipt's hash chain status is broken or has gaps
Alternatives in App and agent security
Something wrong or out of date? Tell us.