Ordewell Desktop · macOS, Windows and Linux

Your coding agents, one plan.

Describe a goal and watch it become a graph of tasks for Claude Code, Codex and the other agents you already use. Review every plan and every diff, or let it run hands-off until the work is merged.

Uses the coding agents you already have, signed in · no API key needed · git optional

A run in progress shown as the plan graph: three tasks in the first stage running on Claude Code, Codex and GitHub Copilot CLI, a test-suite script in the second stage, and a fix task on OpenCode and a read-only review on Claude Code waiting in the third, with the same plan listed in the panel on the right.
A plan of six tasks in three stages, the first three running at once. A capture of the real app on a demo project, with scripted agents standing in for live ones.

Plan

Shape the plan, or let it run.

The planner reads your repository, asks about anything left vague, and writes the work as a graph of tasks in stages, each with its own agent, model and mode.

Rewrite a prompt, move a task to another agent, or ask for a different plan, then approve it. Or turn on autorun and it starts as soon as the planner submits it.

Your first plan in the app →

The plan ready for approval: the planner conversation ends in an approval card, and the Plan panel on the right lists each task with its runner and model.

Run

Claude Code, Codex and OpenCode, side by side.

Each task names the agent and the model that runs it, so one plan can hand its tests to Codex, its docs to a lighter model and a review to Claude Code. Independent tasks run at once, each on its own branch; the ones that depend on them wait.

Open any task to follow its agent at work, or send it a message while it runs.

Running a plan →

A task's chat: GitHub Copilot CLI adding a section on failed sign-in attempts to docs/api.md, with a queued message for it below, while the Plan panel on the right shows three tasks running on Claude Code, Codex and GitHub Copilot CLI and a fix task on OpenCode waiting in a later stage.

Review

Read the diff, or let it merge.

Every change is made on its own branch, where you can read it. Comment on a line and the comment goes back to the task, quoting that file and line.

Merge it onto your branch when it reads right, or turn on automatic merging and Ordewell merges what passes. A conflict stops the merge with your tree untouched.

Reviewing and merging →

The Changes view: a task's changed files beside a side-by-side diff, with a line comment being written to the task.

Git

Git, built in.

Branches, stashes, worktrees and remotes sit in the sidebar. Commit, pull and push without leaving the window, and open any branch to read its history in place.

The branches Ordewell makes are read-only there, and a live run's landing branch is locked until the run ends.

Git in the app →

The Git sidebar: your branches, Ordewell's read-only branches and remote-tracking ones, with a lock on a live run's landing branch.

Watch

A full run, from goal to merge.

A goal becomes a plan, the plan runs on several agents at once, and a task's diff opens for review.

35 seconds of the real app on a demo project. Scripted agents stand in for live ones, so the pace is the script's, not an agent's. Recorded with Ordewell 0.13.1.

Get Ordewell Desktop

Set the goal. Let the graph run.

Step in whenever you want, or not at all.

Free and open source. You need one coding agent, installed and signed in: Claude Code, Codex or one of the others.