Claude Code and Codex together

One tree for agents from every vendor

Download for macOSFree to use. No card.

Mentat runs Claude Code and Codex together in one desktop app, with Grok Build beside them. When one agent launches another vendor's agent from its shell, Mentat nests that run under the session that started it, so a job split across providers reads as one tree.

A Claude Code chat in Mentat. The details rail lists the four agents it started, two of them a Codex run and a Grok Build run, with the checkout they worked in, each turn, and the files they touched.
Close-up of the details rail: under one Claude Code chat, a Codex run and a Grok run listed as its sub-agents and background runs.

01

Agents that start other agents

A run started with claude -p, codex exec or grok -p is matched to the shell command that launched it and listed under that session, next to the lead's own sub-agents.

02

One set of controls, three protocols

Steering, queueing, approvals, plans and goals sit in the same place in every chat, and each is sent through the provider's own mechanism: stream-json for Claude Code, the app-server for Codex, the Agent Client Protocol for Grok Build.

03

Who changed what

Checkouts shows the chat and its agents together, and Files touched gives each file with the lines its agent added and removed.

Questions

Do I need an API key?

No. Mentat runs each command-line tool under the login it already has and never calls a model itself.

Do I have to use all three?

No. Mentat needs one of Claude Code, Codex or Grok Build. Add the others when you want them.

For engineers who stay close to the code

Set your agents in motion

Free to use, with no card. Every new account starts with fourteen days of Pro.

macOS 13 or later, Apple silicon. Windows is next. Git 2.45 or later. At least one of Claude Code, Codex or Grok Build, installed and signed in.