herdr.dev · install · quick start · supported agents · integrations · configuration · socket api
v.0.4.0.mp4
agent multiplexer that lives in your terminal.
workspaces, tabs, panes. mouse-native: click, drag, split. every agent at a glance: blocked, working, done. detach and reattach, agents keep running. no gui app, no electron, no mac-only native wrapper. you see the agent's own terminal, not someone's interpretation of it.
curl -fsSL https://herdr.dev/install.sh | shor download the binary from releases. requires linux or macos.
herdr notifies you when a new version is available. run manually to update:
herdr updateherdrby default herdr launches or attaches to one background session server. ctrl+b q detaches the client. agents keep running. use herdr server stop to stop the default server. use --no-session for the old single-process mode.
named sessions are runtime/socket namespaces for separate persistent herdr servers. they do not replace workspaces; each named session has its own panes, tabs, workspaces, sockets, and session state while sharing the same global config file.
herdr session list
herdr session attach work
herdr session attach side-project
herdr session stop work
herdr session delete side-project- press
ctrl+b, thenshift+nto create a workspace - run an agent in the root pane
- press
ctrl+b, thenwto open workspace navigation - use
ctrl+b, thenvorminusto split panes, orctrl+b, thencto create a new tab - watch the sidebar for blocked, working, and done states
on first run herdr opens a short onboarding flow. after that, restored sessions land in terminal mode; fresh sessions start in navigate mode.
| tmux | gui managers | herdr | |
|---|---|---|---|
| persistent sessions | ✓ | — | ✓ |
| detach / reattach | ✓ | — | ✓ |
| panes, tabs, workspaces | ✓ | ✓ | ✓ |
| agent awareness | — | ✓ | ✓ |
| lives in your terminal | ✓ | — | ✓ |
| real terminal views | ✓ | — | ✓ |
| mouse-native | — | ✓ | ✓ |
| lightweight binary | ✓ | — | ✓ |
| agents can orchestrate | ? | ? | ✓ |
tmux gives you persistence and panes, but it was built before agents existed. gui managers show agent state, but they make you leave your terminal and use their wrapped view. herdr is persistence and awareness in one tool that stays out of your way.
start herdr where the work lives. locally, run herdr. it starts or attaches to the background session automatically, with no socket setup. run your agents, split panes, do your work. press ctrl+b q to detach. close your terminal, close your laptop; your agents keep running. open a new terminal, run herdr, you're back. same session, same panes, same agents.
need to check on your agents from your phone? just ssh in and run herdr. your shell is remote, herdr runs there, and the panes keep running there after detach. any ssh client works. no app to download, no account to create.
ssh you@yourserver
herdr
or attach from your local terminal through ssh without opening a shell first. your local herdr acts as a thin client, connects over ssh, starts or attaches to the remote herdr server, and streams the ui back to your terminal. remote attach uses your local keybindings by default; pass --remote-keybindings server to use the remote server config instead.
herdr --remote workbox
herdr --remote ssh://you@yourserver:2222for repeat targets, use your ssh config:
Host workbox
HostName yourserver
User you
Port 2222same session, same agents, same state.
herdr and herdr --remote attach to the full herdr session ui. herdr agent attach <target> attaches your current terminal directly to one server-owned terminal, like a single-pane terminal attach. herdr terminal attach <terminal_id> does the same by terminal id.
direct attach streams the current rendered terminal state first, then live ansi frames. your input goes straight to that terminal. detach with ctrl+b q; send a literal ctrl+b with ctrl+b ctrl+b. one writable client owns input and resize for a terminal. a second attach fails unless you pass --takeover.
the sidebar shows which agents are blocked, working, or done. workspaces roll up to their most urgent state so you can scan the full list at a glance.
states:
- 🔴 blocked — agent needs input or approval
- 🟡 working — agent is actively running
- 🔵 done — work finished, you have not looked at it yet
- 🟢 idle — done and seen
detection works by reading foreground process and terminal output. zero config, no hooks required. for agents that expose hooks, the socket api integration gives more robust state reporting.
not a gui window, not a web dashboard, not electron. herdr runs inside whatever terminal you already use. single rust binary, no dependencies. works inside tmux.
- workspaces — organized around git repos or folder names, each with its own tabs and panes
- tabs — first-class in the socket api and cli
- mouse-native — click panes/tabs/workspaces/agents, drag borders, select text to copy, right-click menus; not keyboard-only
- notifications — sounds and toasts for background events; tab-aware suppression
- 18 built-in themes — catppuccin, terminal, tokyo night, gruvbox, one, solarized, kanagawa, rosé pine, vesper, and light variants for the main palettes
- session persistence — pane processes survive client detach; sessions restore after full restart
the local unix socket lets agents create workspaces, split panes, spawn helpers, read output, and wait for state changes.
# create a workspace and tab
herdr workspace create --cwd ~/project --label "api"
herdr tab create --label "logs"
# split a pane and run
herdr pane split 1-1 --direction right
herdr pane run 1-2 "npm test"
# wait for a pane-level ui attention state
herdr wait agent-status 1-1 --status done
# read output
herdr pane read 1-2 --source recent --lines 50
# read a rendered ansi snapshot for tui feedback loops
herdr pane read 1-2 --source visible --ansifull reference: socket api and SKILL.md.
automatic detection works out of the box. process name matching plus terminal output heuristics.
| agent | idle / done | working | blocked |
|---|---|---|---|
| pi | ✓ | ✓ | partial |
| claude code | ✓ | ✓ | ✓ |
| codex | ✓ | ✓ | ✓ |
| droid | ✓ | ✓ | ✓ |
| amp | ✓ | ✓ | ✓ |
| opencode | ✓ | ✓ | ✓ |
| grok cli | ✓ | ✓ | ✓ |
| hermes agent | ✓ | ✓ | ✓ |
| cursor agent | ✓ | ✓ | ✓ |
| antigravity cli | ✓ | ✓ | ✓ |
| kimi code cli | ✓ | ✓ | ✓ |
| github copilot cli | ✓ | ✓ | ✓ |
| kiro cli | ✓ | ✓ | — |
detected but not fully tested: gemini cli, cline.
for agents outside the built-in list, herdr still works as a terminal multiplexer with workspaces, panes, and tiling. custom integrations can report agent labels over the socket api. see the socket api docs.
the built-in pi, omp, claude code, codex, opencode, and hermes integrations forward semantic state to herdr over the socket api. install with:
herdr integration install pi
herdr integration install omp
herdr integration install claude
herdr integration install codex
herdr integration install opencode
herdr integration install hermessee the integrations docs for setup details.
press ctrl+b to enter prefix mode. default actions are prefix-first and tmux-like:
| key | action |
|---|---|
prefix+c |
new tab |
prefix+n / prefix+p |
next / previous tab |
prefix+1..9 |
switch tab |
prefix+w |
workspace navigation |
prefix+shift+n |
new workspace |
prefix+shift+g |
new worktree |
prefix+shift+w |
rename workspace |
prefix+shift+d |
close workspace |
prefix+h/j/k/l |
focus pane |
prefix+v / prefix+minus |
split pane |
prefix+x |
close pane |
prefix+b |
toggle sidebar |
prefix+z |
zoom pane |
prefix+r |
resize mode |
prefix+q |
detach |
resize mode: h/l resize width, j/k resize height, esc exit.
custom command keybindings can launch detached shell helpers or temporary panes:
[[keys.command]]
key = "prefix+g"
type = "pane" # "shell" or "pane"
command = "lazygit"if you have old custom keybindings and want the new defaults, run herdr config reset-keys. herdr backs up config.toml, removes only keybinding config, and uses built-in v2 defaults after restart or config reload.
mouse is supported throughout. full reference: configuration docs.
config file: ~/.config/herdr/config.toml
herdr --default-config # print full default configin-app settings screen for theme, sound, and toast preferences. full reference: configuration docs.
herdr writes logs under ~/.config/herdr/.
common files:
~/.config/herdr/herdr.log
~/.config/herdr/herdr-client.log
~/.config/herdr/herdr-server.log
in persistent session mode, the client and server logs are usually the useful ones. logs rotate automatically and keep a few older files like .1 and .2.
for issue reports, include the relevant current log plus rotated siblings if they exist. default logs are metadata-focused and avoid pane contents by default.
use a higher log level only when needed:
HERDR_LOG=herdr=debug herdrfull logging and environment variable details: configuration docs.
- configuration — keybindings, themes, notifications, environment variables
- integrations — pi, omp, claude code, codex, opencode, hermes integrations
SKILL.md— reusable agent skill- socket api — socket protocol and cli reference
if you are an ai agent helping with this repository, read AGENTS.md before making changes and read CONTRIBUTING.md before opening issues or PRs.
git clone https://github.com/ogulcancelik/herdr
cd herdr
cargo build --release
./target/release/herdrherdr provides optional nix flake outputs for users who already use nix. the flake builds herdr from source.
nix run github:ogulcancelik/herdr/v0.x.y
nix build github:ogulcancelik/herdr/v0.x.y
nix develop github:ogulcancelik/herdr/v0.x.yreplace v0.x.y with the latest release tag. you can omit the tag to track master, but release tags are recommended for normal installs.
the flake exposes packages.<system>.default, apps.<system>.default, devShells.<system>.default, and overlays.default.
update through the same nix workflow you used to install herdr. for profile installs, run nix profile list and then nix profile upgrade <index-or-name>. for flake inputs, run nix flake update herdr in your own flake and rebuild.
just test # unit tests
just check # formatting, tests, and maintenance checksagpl-3.0: free to use, modify, and distribute. modified versions must be open-sourced under the same license.
herdr does not require or install terminal keybinds for pi.
ghostty does not ship a default shift+enter=text:\n or shift+enter=text:\x1b\r keybind. if those lines exist in your ghostty config, they were added by user config or another tool, commonly claude code. they collapse shift+enter into legacy bytes, so downstream programs cannot reliably distinguish shift+enter from ctrl+j or alt+enter.
if shift+enter behaves differently in pi inside herdr, first remove those custom terminal keybinds and retest. do not file this as a herdr keyboard encoding bug unless it reproduces with a clean terminal config.
related context: #78, #81, #106, and earendil-works/pi#1872.