ttym

Know which agent needs you. Get back into its terminal from anywhere.

$ curl -fsSL https://raw.githubusercontent.com/esc5221/ttym/master/install.sh | sh

Watch the 69-second film ↓ Internals →

  1. 01Which one needs you
  2. 02Answer in place
  3. 03Same terminal, any device
  4. 04Plans open in the pane
  5. 05Sleep and wake
  6. 06Back after a reboot

Real footage: one Claude Code session on an isolated ttym server, recorded end to end. Nothing is mocked.

01

10 agents running. One is waiting on you.

The work map lists every session with a one-line summary of what it is doing and what it waits on. The one that needs you says so. Click it and you are in its terminal.

# summarize every session that changed
ttym map refresh

02

Jump in. Answer right where it stopped.

The pane in the browser is the terminal itself, not a dashboard. Pick an option on the agent's question card with the arrow keys, type your answer, and it goes back to work. The same session opens in the CLI too.

ttym attach auth-service/limiter

03

Same terminal, on your phone.

One process holds the terminal; the laptop and the phone are both views of it. Type on either and it shows on both. Put ttym on your tailnet and devices signed in to your Tailscale account open it with no extra login.

# or Cloudflare Tunnel + Access
ttym remote tailscale

04

The agent runs ttym open. The plan opens in its pane.

Markdown, HTML, CSV, images and URLs open as tabs on the pane that produced them, not in a browser tab you will lose. Select a path in the terminal and open it the same way.

ttym open PLAN.md rollout.html

05

Idle agents sleep. Same conversation when they wake.

An agent left alone for 30 minutes is stopped and its RAM returned (374 MB in the film). The screen stays. The next keystroke runs claude --resume with the same session id; in the film it was back in 2.5 s.

# or let the 30-minute timer do it
ttym agent sleep :limiter

06

Rebooted? One command brings it back.

After a shutdown every pane comes back with its last screen and scrollback, over a fresh shell. The agent itself is one command away, in the same conversation.

ttym agent resume

Five commands. Then it's yours.

  1. curl -fsSL https://raw.githubusercontent.com/esc5221/ttym/master/install.sh | shDownloads the build for your machine from GitHub Releases, checks its sha256 and installs it to ~/.local/share/ttym, with ttym linked in ~/.local/bin. Needs Node 20 or newer. Read the script first.
  2. ttym service installKeeps the server running: it starts at login and restarts after a crash (launchd on macOS, systemd on Linux). Terminals live in their own processes, so a restart does not close them.
  3. ttym agent install claudeHooks for Claude Code (Codex too): which agent needs you, await, sleep and wake.
  4. ttymOpens a workspace in this terminal. The same sessions are in the browser at http://localhost:7690.
  5. ttym remote tailscaleYour phone, one login. Other ways in.

ttym upgrade swaps in the latest release while your sessions keep running; ttym upgrade --rollback goes back one. Building from source is in the README.

Under the hood

02Sessions outlive the serverA Rust holder per terminal keeps the PTY; a restarted server catches up from a checkpoint plus the missing bytes. 04Agents as functionsttym await returns when the program says it is done: an agent's Stop hook, or a shell's end-of-command mark. 06Remote access, checkedAllow-listed hosts, a login cookie, and identities verified with tailscale whois or an Access JWT.

All seven: holder and server, recovery, the seq, waiting for done, sleep, the remote gate, the repo →