OpenCode Agent Orchestration Kit plugin
A kit that gives OpenCode role-based agents, handoffs, independent review and task state that survives interruptions.
- 148 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Installation edits OpenCode configuration and agents can change files and work on a branch that ends in a pull request.
Install
Manual install
npm install --global opencode-agent-orchestration-kitFirst command of the npm installation section. Then `oak install --dry-run` and `oak install` change the OpenCode configuration.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Installation edits OpenCode configuration and agents can change files and work on a branch that ends in a pull request.
This is third-party code. Review the repository files before installing.
What it does
A configuration kit for OpenCode splits work into roles: routing, research, design, specification, implementation, review and validation. By default the `lead` agent takes a request. It sends small edits straight to implementation and leads large or ambiguous ones through research, a plan, independent review and approval. The bundle has the slash commands `/feature`, `/plan`, `/scope`, `/loop`, `/review`, `/test` and `/code-simplify`, local checklist skills, the `oak` command for installation and runs, and configuration integrity checks. `/loop` works only under an approved task contract and makes at most three iterations per call. `/autonomous` works on a branch and ends at an open pull request through `oak deliver`, with no merge or publish.
Who it is for. Teams and solo developers on OpenCode who want a repeatable process with review instead of scattered prompts.
Good fit when
- A large feature goes from research to review instead of one prompt.
- Each role needs its own model from the providers available in OpenCode.
- Agent configuration has grown and you want to validate it with a command.
- Edit iterations must survive an interrupted session and resume from saved state.
Not a fit when
- Your tool is not OpenCode: the kit is written only for it.
- The task is small and a heavy role-based process would be overkill.
- You want a fully autonomous code factory: the authors explicitly rule this out.
Example request
/plan Add a tag filter to the task list and prepare an implementation plan without changing code.Limitations
Needs OpenCode, Git, Bash, Node.js `^22.9.0 || ^24.0.0` and a model provider configured through `opencode auth login`. OpenCode 1 is the supported target and OpenCode 2 is marked experimental. `oak install` edits the OpenCode configuration directory, so run `oak install --dry-run` first. Open Design, Impeccable and the token panel integrations are optional. Role models come from your OpenCode configuration, so token spend and result quality depend on the providers you pick.
How to disable. Run `./uninstall.sh` from the repository clone or remove the kit files from the OpenCode configuration directory. The scripts back up the previous configuration, so a rollback path exists.
FAQ
What does the npm install do?
It puts only the `oak` and `oc-switch` commands on `PATH` and leaves OpenCode configuration alone. Configuration changes when you run `oak install` explicitly.
Can I try it without a global install?
Yes. Clone the repository and load the kit from its directory: OpenCode configuration stays untouched.
How do I assign models to roles?
Copy `env.example` to `.env`. It holds a shared default model and optional per-role overrides.
Does `/autonomous` merge?
No. It reaches an open pull request through `oak deliver` and stops. Merge, deploy and publish are excluded by the contract.
What does the environment need?
OpenCode on `PATH`, Git, Bash, Node.js `^22.9.0 || ^24.0.0` with npm, and a model provider set up through `opencode auth login`.
Discussion
No questions or reviews yet.
Related
PluginsAn OpenCode plugin that runs a team of parallel agents with separate sessions, a shared task board and messaging
Orchestrator plugin for OpenCode and Codex CLI: multiple agents, hooks, built-in MCPs, ultrawork mode and hash-anchored edits
An OpenCode plugin: an orchestrator hands work to six specialists, each on its own model, and merges their results into one answer
An open source coding agent for the terminal and desktop with build and plan modes
A multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
A layer over coding agents that assembles a temporary specialist team per task inside Claude Code, Codex, Cursor and Gemini
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| OpenCode Agent Orchestration Kit (this card) | no | 148 | |
| OpenCode Swarm | no | 518 | an agent swarm plugin; OAK builds a bounded role-based process with review and durable state |
| oh-my-opencode-slim | no | 10,127 | a slimmed agent bundle; OAK adds command contracts, configuration validation and installation rollback |
README in short
The kit is built around bounded workflows: state is kept in a JSON snapshot and a log, actions are idempotent, and human approval stays mandatory at sensitive steps.