Here’s a question that doesn’t get asked enough in the AI agent space: how do you know the agent actually did what it said it did?
Fetch.ai’s answer arrived on May 12 with the launch of its Agent Execution Verification System, or AEVS. The system generates publicly verifiable, tamper-evident receipts for every tool call an AI agent makes, creating an on-chain audit trail for autonomous software that makes it functionally impossible for an agent to fudge the paperwork on a refund it processed or a payment it approved.
What AEVS actually does
Every time an AI agent executes a task within the Fetch.ai ecosystem, whether that’s processing a refund, approving a payment, or coordinating with another agent, AEVS stamps it with a cryptographic receipt. That receipt is on-chain, meaning anyone can verify it independently.
AEVS is designed to cover the full spectrum of agent actions. Tool calls, inter-agent coordination, payment approvals: each one gets its own tamper-evident record. The system essentially turns every agent interaction into something auditable by default rather than by request.
The market’s response was immediate
FET, the native token that powers agent discovery, deployment, training, and machine learning access across the Fetch.ai network, jumped 3.15% in the 11 hours following the AEVS announcement. The broader crypto market declined by 0.87% over the same window.
FET’s role in the ecosystem extends beyond simple governance or staking. The token facilitates agent discovery and deployment across the Agentverse, which now hosts over 2 million agents. It also underpins AI-to-AI payments alongside USDC, a functionality that Fetch.ai first demonstrated in December 2025 when it announced the world’s first AI-to-AI payment utilizing Visa, USDC, and FET.
Building blocks stacking up
AEVS didn’t arrive in a vacuum. The ASI:One beta launched on November 19, 2025, expanding the platform’s capabilities around multi-agent coordination and interoperability. Then came the AI-to-AI payment milestone in December. Now AEVS adds the verification layer that makes all of it auditable.
The pattern here is worth noting. Each launch addresses a specific gap in the “agents doing real work” pipeline. Coordination came first: getting agents to talk to each other. Payments came second: letting agents transact with each other. Verification comes third: proving those transactions actually happened as reported.
The Agentverse reaching over 2 million agents is its own data point worth sitting with. Verification becomes exponentially more important as the number of agent interactions grows, simply because manual oversight stops being feasible somewhere around the first few thousand agents.