AEVS (Agent Execution Verification System)

AEVS (Agent Execution Verification System), built by agent infrastructure company Fetch.ai, creates tamper-evident, signed receipts documenting every tool call an AI agent makes — capturing the tool name, inputs, outputs, and timing, then cryptographically sealing each record in a verifiable chain. It automatically intercepts agent tool calls without requiring modification of existing tools, uses ECDSA P-256 signatures with KMS backing to seal receipts, and links records in a hash chain that detects if any execution was altered or skipped. Anyone can publicly verify a receipt via API or web explorer using a shared reference ID, without needing access to the underlying codebase. Three visibility modes (public, private, or proof-only) control what gets stored and who can see payloads, and AEVS works as a drop-in with LangChain, MCP, and other agent frameworks. It is open source and pip-installable, currently in beta with no disclosed pricing.