Use cases

Where agents do work someone else needs to check.

Each page covers the problem, what gets recorded, what anyone can check without asking you, and what runs today. Anything not built yet is marked planned.

The use cases

Agent outcomes & evidence reports

A monthly report on what your agents did, judged against criteria fixed in advance.

Agent-to-agent commerce

Two-sided records of work and payment between agents, so closing a job isn't one side's word.

Decentralized inference

Records of who served which request on an open inference network like mesh-llm.

Multi-agent & A2A task evidence

A record of each agent's part in a task handed between agents.

EU AI Act & regulated operations

Records to show what your agents did under the obligations that apply to you.

Governance & oversight

Human approvals and blind human review, kept as records that can be checked.

Software supply chain & CI

One record per pull request: the change, the checks it passed, and who declared they wrote it.

Disputes

Two records, one comparison: agrees, conflicts, or unconfirmed.

A worked example you can run

One buyer, two suppliers, three rounds: who agreed to what?

A buyer agent haggles with two competing supplier agents over three rounds and narrates its reasoning before each move. It picks a winner against a private mandate, and turns down the cheaper final offer because that offer breaks a hard constraint. Each side keeps its own signed record, and a stranger who trusts neither can check the whole transaction offline.

No payment moves, and the agents' reasoning is scripted, with no model call. It is demo code that runs on your machine (bash RUN.sh), not a hosted service.

Built by Jody (independent agentic engineer) · Apache-2.0 · installs capsule-emit from PyPI

Open the repo ↗ Evidence agent ↗
Request a demo → See what runs online