AgentWorkforce/relay

What it solves

Agent Relay addresses the fragmentation and lack of coordination that occurs when multiple AI coding agents work on the same software project. It eliminates the need for manual copy-pasting between agents and human team members, prevents agents from ignoring project rules, and provides a shared history so teams can track decisions and avoid redundant work.

How it works

Agent Relay acts as a coordination layer that connects various agents, tools, and human developers. It provides a messaging infrastructure that allows agents to communicate across different machines, exposes external tools (like GitHub, Linear, and Notion) as a virtual filesystem that agents can interact with using standard commands (ls, cat, grep), and captures session history for searchability. It also allows developers to define multi-step workflows in TypeScript, enabling the enforcement of deterministic rules and human-in-the-loop gates for agent tasks.

Who it’s for

This project is for engineering teams who use AI coding agents and want to move beyond individual, isolated agent workflows toward a collaborative, team-based infrastructure.

Highlights

• Messaging layer for agents to coordinate across different environments and channels.
• Virtual filesystem integration that maps external tools like GitHub and Notion to standard file commands.
• Shared session history to track agent decisions and past work context.
• Workflow engine to enforce deterministic rules and human gates using TypeScript.