Lab / Agent coordination
Relayfox.
File-native coordination for mixed AI agents. Execution stays answerable.
What it is
Relayfox is a coordination layer for mixed-agent AI execution. It uses durable file artefacts — task packets, claim records, lint rules, and result proof — to keep bounded AI work traceable from brief to close.
The problem it solves: AI can make execution feel instant while making drift invisible. A model starts with a clear goal and ends with unreviewable changes, invented checks, or lost context. Relayfox makes the handoff surface explicit.
Current state
Active. The coordination system runs on this site and on Relayfox itself. Task packets, agent bus, lint rules, and result proof are all live. The field report on the first version is in Work.
What is being built next
Tighter agent bus integration. Structured claim diffing. A public schema for task packets so other teams can run the same coordination layer without adopting a new tool.