Plan it. Revise it. Ship it. OpenADE turns agent-assisted coding from a gamble into a workflow. Open source. No login or cloud.
Claude Code and Codex. Same loop, more engines.
Most AI tools fire agents at your code and hope for the best. OpenADE gives you a structured loop that makes agent output predictable.
Describe your task. Get a detailed implementation plan with file-level changes, edge cases, and context-aware suggestions.
Comment on anything. Drill into specifics. Refine the plan until it feels deterministic. Then lock it in.
One click. The agent executes the plan linearly. Automated git snapshots let you rollback to any point.
We use OpenADE 8+ hours a day. Every feature exists because we needed it.
Files, diffs, agent output — everything is commentable. Point the agent at exactly what needs fixing. Type less, iterate faster.
Linear, Notion, Asana, Stripe, Vercel, GitHub and more. Your whole workflow, in the loop, out of the box.
Your code never leaves your machine. No cloud. No syncing. Works without internet. Optional anonymous telemetry, easily disabled.
Every execution creates a patch snapshot. Rollback to any point with a click. Never lose work.
Diff viewer, file browser, fast search, full terminal, and process manager. All without leaving the app.
Isolate experiments in git worktrees. Keep your main branch clean while agents explore possibilities.
Run multiple agents in parallel — even across providers — then reconcile into one plan. Ensemble and Cross-Review strategies built in.
No separate billing. No per-seat licensing. Works with your existing Claude Code or Codex subscription.
Free and open source. Always.
Windows support is early — expect rough edges. For a smoother experience, try the Linux build via WSL.