Approval where it counts
Files, commands, cloud resources, remote devices, and spend wait for your review. You see the action before it happens.
The local-first control plane for AI work
Let agents work.
Keep control.
OpenToad coordinates the agent CLIs you already use, carries workspace context forward, and puts a visible approval boundary around consequential work.
Runs on your machine first · cloud and remote resources only when you allow them
LIVE SESSIONS · steer agent work in contextQUEUED JOBS · continue safely in the backgroundRESOURCE POOL · local first, then enrolled machines
Built for real work
One place to run
the work.
One clear boundary
around the risk.
Your agents, in one room
Start Claude, Codex, Cursor, or Devin on a project, watch tool activity live, and steer the same provider conversation when the work needs direction.
Context that stays useful
Feed files, preserve project decisions, and inspect or delete durable memory. Each run starts closer to the actual work.
The operating loop
From goal
to finished work.
01Describe
Give OpenToad the goal in plain language, with the workspace and constraints that matter.
02Route & steer
Run interactively or queue the work; OpenToad prefers your local machine and explains every routing choice.
03Review & approve
Approve the moments that deserve a human decision, then keep the complete audit trail.
Start with the control plane