Evidence Relay

Inspiration

Coding agents can make changes quickly, but a handoff is only trustworthy when the next operator can see who changed which files, what commands were run, and what evidence passed. We wanted a lightweight, local-first way to turn that invisible coordination work into an auditable record.

What it does

Evidence Relay is a coordination ledger for coding-agent teams. It lets an operator claim a file scope, record decisions and test evidence, surface conflicting active claims, and export a portable handoff note. The dashboard makes the current state legible at a glance while the CLI keeps the workflow fast in a terminal.

How we built it

The project is a Python CLI with a small static dashboard. The ledger is intentionally local and file-backed so it can be inspected, versioned, and moved with the repository. We used Codex with GPT-5.6 to design the workflow, build the command paths and dashboard, write documentation, create the demo, and iterate with tests.

Challenges we ran into

The hard part was making coordination useful without making it bureaucratic. We focused on a tiny set of primitives: claim, evidence, verify, release, and handoff. We also made conflicts explicit, so agents do not silently edit the same file.

What is next

Next we want integrations for common agent runtimes and CI, richer evidence adapters, and a shared-team mode that still preserves the same audit trail.

Built With

• codex
• html
• javascript
• python