StarAgent CLI
Launcher and dashboard for long Codex, Claude Code and OpenCode sessions: tmux, a web panel and Tailscale links.
- 63 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The panel gives web access to terminals where agents run commands and the hub opens a port on the network.
Install
- Works in
Manual install
pip install -e .The README command runs in a cloned repository directory; then `staragent` opens the panel.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The panel gives web access to terminals where agents run commands and the hub opens a port on the network.
This is third-party code. Review the repository files before installing.
What it does
StarAgent runs console agents in persistent terminal sessions and gathers them in one web panel. The `staragent` command opens a local panel at `127.0.0.1:8080` and `staragent hub` starts a hub that joins several machines. Sessions live in tmux on Linux and macOS and in ConPTY on Windows, so `/resume` is not needed. Each session has a chat, a live terminal and a file explorer, and the list switches from a phone. Sign-in and credential sources are chosen on the agents page, and models and reasoning levels are set as defaults without editing vendor configuration. Tailscale links the machines. The repository also has the `staragent-use` and `staragent-simplify` skills for operating the station.
Who it is for. Developers who run several agents across repositories and want to watch them from a phone and other machines.
Good fit when
- Three tasks run in parallel in different folders and one status screen is needed.
- An agent session must survive a closed terminal and return without `/resume`.
- An agent works on a server and is easier to check from a laptop or phone.
- Each machine gets a default model and reasoning level.
Not a fit when
- There is a single agent in a single terminal.
- The server must not expose a panel on the network or run Tailscale.
- A stable product is needed: the author calls the project personal and rough.
Example request
Start StarAgent, create a Claude Code session for the api folder and show how to open its terminal from a phone.Limitations
The project is personal, changing actively, and the author says it was mostly built with vibe coding, so bugs are possible. The panel gives access to terminals with agents: run the hub only on a trusted network, take the token from the `staragent hub` output and do not expose the port. Tailscale is needed to link machines. Installation is from source with `pip install -e .` and Linux and macOS need tmux. The author promises a stable release later.
How to disable. Stop the `staragent-launcher`, `staragent-hub` and `staragent-node` tmux sessions, then remove the package with `pip uninstall` and the project folder. Leaving Tailscale is done separately.
FAQ
How does Launcher differ from Hub?
Launcher runs on one machine. Hub gathers several nodes, each running `staragent node-ts` or `staragent node` to hold sessions.
Which agents are supported?
Codex, Claude Code, OpenCode and a plain shell. The agents page shows availability, sign-in, skills and install or update paths.
Is there a desktop app?
Yes, prebuilt `.dmg` packages for macOS and an AppImage for Linux, while Windows uses ConPTY. Packages ship their own Python runtime.
How to verify a node?
On the hub machine run `staragent verify-node <node-address>`, then add host and port in the panel, for example `100.x.x.x` and `8081`.
Is Tailscale required?
Recommended for linking machines. On a LAN `staragent node` is enough, without `tailscale serve`.
Discussion
No questions or reviews yet.
Related
CLI toolsA tmux-based TUI that runs Claude Code, Codex and other agents in one list: status, quick prompts, worktrees and diff review
A local dashboard for running many agents across machines: memory, code graph and agent-to-agent messaging
Local meeting rooms for Codex and Claude Code agents to talk to each other while a person reads and posts from a browser.
A CLI with a TUI that finds Claude Code, Codex and Gemini CLI sessions and continues them in another agent as native ones
A terminal session manager for Claude Code, Codex, Gemini CLI, OpenCode and other agents: one list with statuses, git worktrees and control from a phone
A TypeScript terminal agent: streaming REPL, OpenAI-compatible and Anthropic models, plan mode and edit rollback.
README in short
The README explains the panel, nodes and sessions, with separate documents on the desktop app, the hub, Lark integration and the session model.