bohay: mission control for your AI coding agents

4 min read Original article ↗

Next-Gen mission control
for your AI agents

One place to run, watch, resume, and orchestrate every coding agent.

curl -fsSL https://bohay.dev/install.sh | sh

or brew install RizRiyz/bohay/bohay · cargo install bohay · Windows & more

bohay · sudos ● live demo · click around

+

⌃Space prefix · copilot is blocked · click its pane to approve

What it does

Everything your agents need, in one place

A terminal that watches your agents, reads your repo, and ships your code, without another window.

Watch

Status you can trust

Which agents are working, which are blocked on you, which have finished, and where each one lives. An agent only counts as working on positive evidence, so a launching CLI or your own typing never reads as busy.

Agent detection →

Supervise

Cost and context, per project

One tab lists every agent in a folder with how full its context window is, what the session has cost, and the project total. Answer a blocked agent inline, or jump to its pane. Token figures come from Claude's own session files today.

Mission Control →

Ship

Git without leaving

Click a branch and a tab opens on it: commits, branch flow, branches, pull requests, issues and a repo overview. It drives git and the gh CLI you already have, so it never disagrees with your shell.

The git tab →

Read

Read what agents touch

A file tree tinted by git status, and a viewer that refreshes itself as the file changes on disk. Read it in place, or send it to vim, nano, or whatever $EDITOR is. Modules can add their own panel below it.

Files and the viewer →

Resume

Agents pick up where they stopped

bohay finds each agent's own session store on disk and runs its resume command for you. Nothing to configure, no session flag to paste. Works for claude, codex, copilot, opencode, kimi, grok and pi.

Session resume →

Fork

Branch a session without losing it

Right-click a pane and choose Fork to New Pane. The session branches into a pane beside it, carrying the whole conversation, while the original keeps running. Try an approach without gambling the context you built. Also ⌃Space f.

Forking a session →

Reach

Drive another machine

Attach to a session running on another box over plain SSH. No port forwarding, no daemon to install. Only the changed cells cross the wire, about 22 bytes for a keystroke, so it types like it is local.

Remote attach →

Mobile

Your session, from a phone

SSH in from any terminal app (Termius, Blink, Mosh) and run bohay. Under 50 columns it goes compact on its own: sidebars away, one pane full screen, and a switcher built for thumbs. It is the same session your laptop left running.

Working remotely →

Notch

The same team, in the notch

bohay-notch mirrors every agent into a native macOS panel that drops out of the MacBook notch. Watch what each one is doing, answer a request or grant permission, and jump straight to its pane, without leaving the app you are in.

How the notch app works →

Also: a worktree per branch · modules in any language · remappable keys · movable panels and themes

Several agents. One repo.
Zero conflicts.

Add tasks with dependencies to the ◇ orch board. Each started task gets an isolated git worktree and its own pane. Path leases stop overlapping edits from even being assigned. A quality gate (your test command) must pass before a task counts as done, and finished branches merge through a dedicated worktree, so your checkout is never touched.

Read the orchestration guide →

Works with the agents you already use

Claude Code Claude Code

Codex Codex

Gemini CLI Gemini

Aider Aider

opencode opencode

GithubCopilot Copilot

Kimi Kimi

Qwen Qwen

Kiro Kiro

Cursor Cursor

Amp Amp

Droid

Grok Grok

Pi Pi

Live status for every one of them, zero setup. The agents also resume your sessions after a restart, and most report precise events through the optional hook. Full matrix →

Open source · AGPL-3.0 licensed

If bohay saves you time, give it a star

bohay is free and fully open source. A star helps more people running AI agents find it, and you are welcome to read the code, open an issue, or send a pull request any time.

Star on GitHub Browse the code