OrbitDock

OrbitDock

Run, review, and orchestrate AI coding agents from anywhere, including your phone.

Why

I don't write code anymore (pretty much) — agents do. My job is reviewing, managing, and
providing guidance at the right time. But with multiple repos and a bunch of agents running
across all of them, keeping track of it all was chaos. OrbitDock is how I wrangle that.

A Rust server sits at the center. It creates and runs both Claude Code and Codex sessions
directly — Codex via embedded codex-core, Claude via the CLI. It also picks up existing Claude
Code terminal sessions through hooks. Run a server on your laptop, your work machine, a VPS,
wherever your agents are. The macOS and iOS apps connect to all of them at once, so you get one
unified view no matter where your sessions are running or what device you're on.

What it does

Mission Control. Point it at a Linear project and agents start working issues on their
own. OrbitDock pulls eligible issues, creates per-issue git worktrees, picks the right
provider, and dispatches agents with a prompt template you control. Concurrency limits,
retry logic, provider failover — all configurable through a repo-local MISSION.md. A
dedicated dashboard shows the full pipeline. Hands-off orchestration until you need to
step in.

Bidirectional, not just passive. Hooks tell you what's happening. OrbitDock lets you act on
it. Send messages, steer mid-turn, fork conversations, roll back turns — all from the app. You
can also take over a passive hook-monitored session and convert it to full direct control
without interrupting it.

Code review that feeds back. A magit-style diff view shows exactly what the agent changed.
Add inline comments on specific lines, then send them as structured feedback. The agent gets
your context and corrects…