agent-manager CLI
A tmux-based TUI that runs Claude Code, Codex and other agents in one list: status, quick prompts, worktrees and diff review
- 616 ★ on the repository
- Last commit
Broad access
Why this level
- The manager launches agents and shells on your machine, including in separate git worktrees
- Through MCP, agents gain the ability to launch other agents and message them
- The install script downloads and installs a binary from a release
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agent-managerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The manager launches agents and shells on your machine, including in separate git worktrees
- Through MCP, agents gain the ability to launch other agents and message them
- The install script downloads and installs a binary from a release
This is third-party code. Review the repository files before installing.
What it does
agent-manager is a local terminal app that launches your installed agent CLIs as they are, with your login, subscription, config files and MCP servers. It supports Claude Code, Codex, OpenCode, Grok, Gemini CLI, Antigravity CLI, Pi, Command Code, Hermes Agent, Muse Code, Oh My Pi and Cursor CLI. Each agent runs in its own persistent tmux session and keeps going after the manager is closed. All sessions appear in one list with live status, grouped into a project tree. The space key sends a prompt straight into a session's pane or spawns a new agent in the selected group. The v key revives a stopped session on its own conversation, and f creates a separate fork. Ctrl+R opens a full-file diff of the agent's changes: line comments sent with the C key go back to the agent's pane as one prompt. A session can start in its own git worktree, and the model, reasoning effort and profile are picked at launch. An agent can spawn another agent, message it and wait until it finishes: every MCP-capable session gets these tools on launch. Terminals for builds and git live in the same list. Cost tracking is not implemented yet.
Who it is for. For developers who keep several agents in the terminal at once and want to see which one has finished and which is waiting for an answer.
Good fit when
- Claude Code works on one task and Codex on another, and you want one status list instead of terminal tabs
- A stuck agent needs an answer without entering its session: space sends the prompt straight into its pane
- An agent's result must be reviewed line by line: Ctrl+R opens the diff and the comments go back as one message
- Each task needs its own branch: the session starts in a dedicated git worktree
- A session stopped or you need memory back: the x key frees resources and v returns the session to the same conversation
Not a fit when
- You need token and cost accounting per session: the project marks it as not available yet
- You want a graphical interface: the app runs in the terminal
- Windows has neither psmux nor WSL2: native operation is built on psmux
- You need a model gateway that swaps keys: the manager launches your CLIs as they are and substitutes nothing
Example request
Start three Claude Code sessions in agent-manager for the three tasks on the list, each in its own worktree, and tell me when they finishLimitations
The app runs on macOS, Linux, natively on Windows with psmux and inside WSL2. It needs tmux 3.1 or newer and git, and the installer offers to install them. Hermes, Antigravity, Oh My Pi and Cursor CLI do not support forks.
The install script on the main branch is not in checksums.txt, so read it before running. Verification with gh attestation verify is described for release files. The Scoop package is updated separately and may lag behind the latest release.
How to disable. Agent sessions live on a separate tmux server named agentmgr, and tmux -L agentmgr kill-server stops them all. Remove the app with the package manager you used to install it, for example brew uninstall agent-manager.
FAQ
How do I install agent-manager on Mac or Linux?
With brew install agent-manager: Homebrew installs tmux too if it is missing. There is also a prebuilt binary through brew install --cask yoanwai/tap/agent-manager and an install.sh script that downloads the release, verifies checksums and puts the file in ~/.local/bin.
Which agents does agent-manager work with?
Claude Code, Codex, OpenCode, Grok Build, Gemini CLI, Antigravity CLI, Pi, Command Code, Hermes Agent, Muse Code, Oh My Pi and Cursor CLI. The launch, revive and status rules for each CLI ship inside the binary, so an app update brings their current versions.
How do I review and comment on an agent's changes?
Select a session and press Ctrl+R: a full-file diff opens with syntax highlighting. The c key comments on a line, and the C key sends all comments to the agent's pane as one prompt; sent comments stay visible as open or handled.
How does one agent launch another?
Every MCP-capable session gets the manager's tools on launch: create a session, send a message and wait until it finishes. Saved named instructions are available through create_session, spawn --saved and send --saved.
Do I need separate keys or a subscription?
No. The manager launches the CLIs you already have installed, and they use your login, subscription, config files and MCP servers.
Discussion
No questions or reviews yet.
Related
CLI toolsOpen-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
An engineering system for coding agents: 68 agents, 292 skills, hooks, memory, rules and the AgentShield security scanner
DeepSeek Harness is an open-source AI agent from DeepSeek where every feature is a plugin. It runs in the browser, on macOS and on Windows.
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Ollama runs open models on your own machine and serves them to agents through a local API compatible with OpenAI and Anthropic.
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| agent-manager (this entry) | no | 616 | |
| Agent Deck | no | 1,147 | also a shared session list with statuses and worktrees, plus control from a phone |
| Agent of Empires | no | 3,377 | isolated tmux sessions with a TUI and an extra web dashboard |
| Claude Squad | no | 8,656 | built for Claude Code, Codex, Gemini CLI and Aider in tmux sessions and worktrees |
README in short
The description shows in animations how five prompts go to five agents, how the mouse works and what the full-screen list looks like. Next are the list of supported CLIs, install methods for macOS, Linux and Windows with checksum and provenance verification, a short key table and worktree launch. The detailed reference, settings and install documentation live in the docs directory.