Relay

Coding agents

Claude Code, Codex, Grok—and the terminals you already use.

Relay does not make you choose a single provider. It gives each coding agent its own live terminal node, account, role, and place in the work so you can use the right agent for each part of a build.

In practice

Built for the way agents actually work.

01

Bring your own agent

Run Claude Code, Codex, Grok, Antigravity, OpenCode, or a plain shell. Relay works around the CLIs and accounts you already use.

02

Separate accounts per node

Two nodes can run the same agent with different logins: the CLI's own login, a second login, an API key, or an Anthropic-compatible endpoint such as z.ai. Keep work and personal accounts apart.

03

Reusable roles

Describe how an agent should work once in Markdown, then assign that role to any node. Builder and Reviewer ship built in; write your own tester, researcher or docs writer beside them.

04

A visible multi-agent workflow

See what every agent is doing, where a hand-off leads, and which terminal needs a decision. You keep control without collapsing the work into one chatbot conversation.

Your next build

Put your agents on the same canvas.

Get Relay · $10