@ordewell/cli
v0.11.3
Published
Ordewell CLI and TUI — plan, run and verify coding-agent tasks from the terminal.
Maintainers
Readme
@ordewell/cli
The command line and terminal UI for Ordewell, task orchestration for coding agents. One goal becomes an editable plan with a runner and model assigned to each AI task. With worktree isolation enabled, code tasks run in separate worktrees; ops and script tasks run at the workspace root. Ordewell records task outcomes and combines code changes for review and merging.
This README describes 0.11.3. See the published releases for versions available to install and their release notes.
Installation
npm install -g @ordewell/cli # or: npm install -g ordewellRequirements: Node.js 20.3 or newer, at least one of Claude Code, Codex or OpenCode, and git for task isolation. Linux, macOS and Windows are supported; on Windows, run the terminal UI under WSL.
Usage
Run ordewell in your project to open the terminal UI: the conversation on the
left, the live plan on the right (or beneath it, below 70 columns). The first run lets you pick a planner and runners
with /planner (Harness › Model › Effort, each step saved as you go) and /runners, and no API key is needed if you plan with a coding
agent you already use.
By default, submitted build plans wait for your approval. Autorun can start them
automatically; research tasks and consented script tasks can start during planning.
Automatic merging is optional, and an Act supervisor can also merge.
The welcome screen lists the repositories and branches code tasks will use, and
whether they get isolated worktrees.
/settings walks autonomy, supervisor, merge automatically, autorun and parallel
(research uses the same Parallel slots) in one panel and saves only what you changed. /planner walks harness, model and
effort in one flow, so there is no separate model or effort command. Press ? in
the plan pane for the full key sheet; the footer shows only the keys that apply
to the selected task.
Most commands in the terminal UI are also subcommands. The exceptions: /help
is ordewell --help; /quit, /new (every ordewell plan starts a fresh
session) and /save (plans persist as they change) have no subcommand; /load
and /delete are ordewell sessions load|delete; /open opens a task's view
in the full-screen terminal, where a shell has no view to switch to
(ordewell checkpoint answers from one); /mouse configures that terminal; and
/settings is a panel over headless commands that remain for scripts
(ordewell auto, supervisor, automerge, autorun and parallel);
ordewell supervisor reads and ordewell trailer have no panel row. /planner has its own ordewell planner, and its model and effort
steps are ordewell model and ordewell planner-effort.
export AI_PROVIDER=claude-code # plan with Claude Code, Codex or OpenCode
ordewell plan --goal "Add rate limiting to the public API"
ordewell run
ordewell handoff review # read the diff
ordewell handoff merge # bring it onto your branchRun ordewell --help for the full list, or ordewell setup to write a
starter .env file.
Type /name in the conversation to load a skill, and use ordewell skills to
list the ones a workspace sees and ordewell task-skills to attach them to a
task. Writing skills covers the format.
This package also installs @ordewell/web, the local API server that the CLI and
terminal UI talk to over 127.0.0.1. It starts on demand.
Documentation
Full documentation is at ordewell.ai/docs, and the source is at github.com/ordewell/ordewell.
