#clawsh docs

clawsh lets you see and drive every Claude Code session on every machine in your LAN from one terminal.

It is one binary with two roles:

• agent runs on each machine that hosts sessions. It owns claude processes inside PTYs (ConPTY on Windows), keeps them alive while nobody is watching, and serves a small REST + WebSocket API.
• hub is the set of commands you type (hosts, ls, new, attach, kill, and web for a browser UI) from whichever terminal you're in. It reads a list of agents from hosts.toml.

$ clawsh ls
TARGET         NAME      STATUS   VIEWERS  AGE    DIR
desk/815856a6  api       running  0        2h14m  /home/alvaro/src/api
laptop/3fa0c1  frontend  running  1        12m    C:\src\frontend

#Pages

• Installation: install script, npm, prebuilt binaries or from source
• Quick start: agents, hosts.toml, your first session
• Commands: every subcommand and flag
• Web UI: clawsh web, the browser dashboard
• Configuration: hosts.toml, tokens, file locations
• How it works: sessions, scrollback, attach
• Protocol: the agent's HTTP and WebSocket API
• Claude Code plugin: auto-start the agent, list every session (experimental)
• Security: what a token grants and how to keep it safe
• Development: building, testing, contributing
• Roadmap and limits: what works today and what's next

Edit this page on GitHub