claude-code-orchestra workflow
A template sets up a multi-agent environment in a repository: Claude Code conducts Sonnet and Opus subagents and calls Codex CLI.
- 198 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Hooks and skills run scripts and commands on your machine, while subagents and Codex CLI edit project files.
- The installer changes `CLAUDE.md`, `AGENTS.md` and the `.claude/` directory in the repository.
Install
- Works in
Manual install
template_dir="$(mktemp -d)"
git clone --depth 1 https://github.com/DeL-TaiseiOzaki/claude-code-orchestra.git "$template_dir"
bash "$template_dir/scripts/install.sh" . && rm -rf "$template_dir"Run it at the root of an existing repository; for a new project use the Use this template button on GitHub. Then start `claude` and run `/init`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Hooks and skills run scripts and commands on your machine, while subagents and Codex CLI edit project files.
- The installer changes `CLAUDE.md`, `AGENTS.md` and the `.claude/` directory in the repository.
This is third-party code. Review the repository files before installing.
What it does
The installer drops a `.claude/` directory of agents, rules, hooks and skills into the project, plus a root `AGENTS.md` with the orchestration policy. Claude Code runs the conversation, hands routine implementation to a Sonnet subagent, research and hard tasks to an Opus subagent, and passes planning and complex code to Codex CLI. The kit includes `plan`, `tdd`, `spike`, `troubleshoot`, `simplify`, `research-lib` and other skills, and hooks that run a linter on save and checks after tests. A GitHub template covers a new project, and `scripts/install.sh` handles an existing one without overwriting your files.
Who it is for. Developers already paying for Claude Code and Codex who want a ready division of roles between models.
Good fit when
- A large task is split: research goes to Opus, edits to Sonnet and the plan is agreed with Codex.
- Shared rules, hooks and skills are installed into an existing repository with `install.sh`.
- Tests fail and the error is automatically sent to Codex for analysis.
- The project needs state kept in `.claude/STATE.md` so a new session resumes where the last stopped.
Not a fit when
- You work only in Claude Code with no Codex CLI and no wish to install a second tool.
- You need one small skill, not a whole environment with hooks.
- The repository is shared and foreign hooks in `.claude/` are not acceptable.
Example request
Plan the payment module migration through Codex, research libraries with an Opus subagent and hand the edits to a Sonnet subagent.Limitations
The full version requires installed and authenticated `claude` and `codex`. A single-Claude-Code variant lives on the `claude-only` branch. The Codex model is set in `.claude/settings.json` and `.codex/config.toml` and must be bumped by hand on updates. Hooks run Python scripts on your machine, and `update.sh` never merges `settings.json` automatically. Python hooks and scripts run without separate confirmation, so review `.claude/hooks/` and `.claude/settings.json` before installing.
How to disable. Delete the `.claude/` and `.codex/` directories and the `AGENTS.md` and `.claude/STATE.md` files. If the installer moved your `CLAUDE.md` aside, restore it from `CLAUDE.md.pre-agents-md.bak`.
FAQ
How do I install the template into an existing project?
Clone the repository into a temporary folder and run `bash "$template_dir/scripts/install.sh" .` at the project root. Path conflicts stop the install, and you can retry with `--force` after reviewing the report.
What does the installer do with my `CLAUDE.md`?
It moves it to `CLAUDE.md.pre-agents-md.bak`, because Claude Code reads `AGENTS.md` only when no `CLAUDE.md` shadows it. Merged content goes to `.claude/STATE.md`.
Can it run without Codex?
Yes, the `claude-only` branch replaces Codex roles with Opus subagents and only `claude` is required.
Which commands does the Codex plugin for Claude Code add?
The optional plugin adds `/codex:review`, `/codex:adversarial-review`, `/codex:rescue`, `/codex:status`, `/codex:result` and `/codex:cancel`.
Which skills ship with the template?
`plan`, `tdd`, `spike`, `troubleshoot`, `simplify`, `research-lib`, `design-tracker`, `feature`, `checkpointing`, `team-execute` and other folders under `.claude/skills/`.
Discussion
No questions or reviews yet.
Related
WorkflowsMulti-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
A Claude Code workspace folder: hooks with a local Ollama detect the project type, pick skills and remind you to update the docs
Workflows for Claude Code: /do, /omo, /bmad-pilot and more commands, with Codex, Gemini or OpenCode writing the code through codeagent-wrapper
A Claude Code plugin that hands coding tasks to Codex agents, monitors them and verifies the results independently
A skill that turns the main Claude Code session into a dispatcher: tasks go to background subagents and the chat never blocks.
A template for Claude Code projects: parallel agent swarms, workflow skills, safety hooks and rules in one install.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| claude-code-orchestra (this card) | no | 198 | |
| Claw Orchestrator | no | 606 | an orchestrator that runs several agents; this is a ready in-repository environment with roles, hooks and skills |
| multi-agent-starter | no | 106 | a starter for several agents; Claude Code Orchestra assigns models to roles and connects Codex CLI |
README in short
The repository is designed as a project template: three tiers of executors, delegation rules and standard work loops. The installer preserves project-owned files and installs safely.