l0ng-ai
d7dad45fa8
* fix(mobile): bump iroh to 1.3.0 so a stuck relay cannot stall direct dials iroh 1.2.0 sends a connection's first datagrams to every known path one after another inside the remote-state actor, awaiting each. When the relay is unreachable its send queue fills and the actor blocks, so handshake packets for a direct path that does work (a mesh VPN address, a public IPv6) queue behind it and the dial times out. 1.3.0 sends to all paths concurrently with a bounded wait (n0-computer/iroh#4512). The desktop workspace was already on 1.3.0; the app has its own lockfile and was left behind. * fix(gateway): reach the relay through a proxy when that is the only way out iroh dials its relay with its own resolver and its own TCP, ignoring the system proxy. On a machine whose network only works through a local proxy (Clash and the like, system-proxy or TUN mode alike) that dial never succeeds, and nothing says so: phones on the same network still connect, while a phone on cellular times out, because without a relay nothing coordinates hole punching and the home router drops unsolicited inbound packets. The gateway now lists the ways out it knows of, most likely first: tty7's own http_proxy setting, the system proxy, the environment's, then direct. It starts on the first without waiting, and when the relay stays unreachable for 10s it tries the others on a throwaway endpoint, switching to the first that reaches a relay (same key, same port, so pairing codes keep working). While none does, it looks again every minute. A pairing code now names the relay only once it is actually connected; before, it named whichever relay iroh picked by latency, reachable or not. The platform proxy readers in daemon::install::proxy now also hand back the proxy as a URL, for a client that is not ureq. Claude-Session: https://claude.ai/code/session_018wE9ZRyxgWW2f55VSy9FvZ
tty7
The terminal that outlives its window.
Pure Rust · GPU rendering on Zed's gpui · VT core from Alacritty
English · 简体中文
Most terminals tie your shells to a window. Close the window and the shells die with it. So people run tmux inside their terminal to keep them alive, which works, but it means learning a second set of keys and keeping a second config file, just to get around the window.
tty7 takes a simpler approach. The shells belong to a server running in the background, and the window only shows them. Close the window and nothing happens to the shells. Most of what's interesting about tty7 follows from that one decision.
Quitting doesn't kill anything
Quit tty7 and every shell keeps running. Your build doesn't notice.
A reboot is different. The processes are gone, because that's what a reboot does. But each pane comes back with its layout and the last of what was on screen, and supported agents pick up the same conversation where they left off.
You don't need tmux for any of this, and there's nothing to configure.
Agents that run agents
A lot of people now run several coding agents at once, in several repos. Which means spending the day switching between windows to see which one finished and which one is waiting for an answer. That's not work a person should have to do.
tty7 recognizes 26 coding CLIs, including Claude Code, Codex, Gemini, Cursor, and OpenCode, and puts all of them in one sidebar: whether each is working or waiting, a notification when one needs you, and its branch and diff. You can see at a glance which one wants you.
Once an agent's status is something a program can ask about, the one asking doesn't have to be you. It can be another agent. The whole loop is four commands, with no framework, and it works whether or not the GUI is open:
PANE=$(tty7 split --v) # give the worker a pane
tty7 send "$PANE" 'claude "add tests for the parser"' --enter
tty7 wait "$PANE" --until waiting,done --changed --timeout 600 # until it's done or stuck
tty7 capture "$PANE" --plain # read what happened
tty7 doesn't wrap the agents or sit between you and them. The agent you start is the real one, running in an ordinary PTY. Simple things are more likely to keep working.
→ Orchestrating agents · agent skill · full support matrix
Remote is the same thing
A remote workspace is the same server, running on the other machine. The window here just shows it. Tabs, panes, the file tree, git, and diffs all live over there, and no files get synced. Connect from a different laptop and everything is where you left it.
So remote isn't really a separate feature. It's the local case with the server somewhere else.
The SSH client is tty7's own, written in Rust, so it doesn't depend on the
system's ssh. It has profiles with keychain secrets, jump hosts, SFTP, and
automatic port forwarding. The remote tty7-server installs once and doesn't
need root.
Speed
Speed isn't the point of tty7, but it shouldn't be a weakness either, and it
turns out not to be. cat on an 11 MB file takes 95 ms; the next fastest
terminal we measured takes 179. DOOM-fire runs at 888 fps; the next best manages
617. The method and scripts are public, and one command reproduces them.
| tty7 | Alacritty | Ghostty | Kitty | |
|---|---|---|---|---|
Plaintext I/O — 11 MB cat (lower = better) |
95 ms | 239 ms | 179 ms | 185 ms |
| DOOM-fire frame rate (higher = better) | 888 fps | 485 fps | 552 fps | 617 fps |
| Cold-launch memory | 116 MB¹ | 105 MB | 128 MB | 130 MB |
Apple M1 Pro, macOS 26.3.1, 155×40 grid, five-run averages, 2026-07-04. ¹ GUI 105 MB + the persistent server 11 MB.
Methodology and one-command reproduction: scripts/bench/.
Everything else
The rest is details, but details are most of what makes a tool pleasant to use. The prompt has suggestions from your history, tab completion that explains each option, syntax highlighting, and real multi-line editing, and ⌃ R searches your history. ⌘ P searches everything. ⌘ J opens a side panel with processes, ports, files, changes, and GitHub issues and PRs. Shell integration works in zsh, bash, fish, PowerShell, and WSL without installing anything.
→ Full documentation · keyboard shortcuts · config.json · CLI reference
Install
Native builds on Releases:
| macOS | .dmg for Apple silicon or Intel — drag into Applications |
| Windows | -setup.exe, or the portable .zip |
| Linux | .AppImage — chmod +x and run; X11/Wayland libraries bundled |
Give your agents the CLI skill:
npx skills add l0ng-ai/tty7 # install
npx skills update tty7 # update later
Supported agents
Detection is free: brand avatar, branch + diff, tab title.
Status takes one click under Settings → Integrations to install that agent's hook,
and brings the status dot, notifications, the tray icon, tty7 wait, and resume
after a reboot. Fork needs both — the agent's own fork command, and the hook
that tells tty7 which session to fork. Past sessions are read from the agent's
own history files and listed in Search Everywhere, where ⏎ resumes one.
The full support matrix
| Agent | Detected | Status · resume | Fork | Past sessions |
|---|---|---|---|---|
| Claude Code | ✓ | ✓ | ✓ | ✓ |
| Codex | ✓ | ✓ | ✓ | ✓ |
| TraeCode | ✓ | ✓ | ✓ | |
| Grok | ✓ | ✓ | ✓ | |
| OpenCode | ✓ | ✓ | ✓ | ✓ |
| Oh My Pi | ✓ | ✓ | ✓ | ✓ |
| Prime Agent | ✓ | ✓ | ✓ | |
| Droid | ✓ | ✓ | ✓ | ✓ |
| Qwen Code | ✓ | ✓ | ✓ | ✓ |
| Goose | ✓ | ✓ | ✓ | |
| Qoder CLI | ✓ | ✓ | ✓ | ✓ |
| Qoder CN CLI | ✓ | ✓ | ✓ | ✓ |
| CodeBuddy | ✓ | ✓ | ✓ | ✓ |
| Gemini | ✓ | ✓ | ✓ | |
| Copilot | ✓ | ✓ | ✓ | |
| Kimi Code | ✓ | ✓ | ✓ | |
| Pi | ✓ | ✓ | ✓ | |
| Crush | ✓ | ✓ | ||
| Antigravity | ✓ | ✓ | ||
| Cursor | ✓ | ✓ | ✓ | |
| Aider | ✓ | |||
| Amp | ✓ | |||
| Auggie | ✓ | |||
| Hermes | ✓ | |||
| Vibe | ✓ | |||
| Empryo | ✓ |
Built on gpui and alacritty_terminal · Apache-2.0 · Discord · Changelog