Define once. Let your agents handle the rest.
The open-source AI Agent first work orchestration framework.
Compose operations into pipelines, plug in any AI agent, and automate any workflow — code quality, data processing, or your own domain.
Documentation · Contributing · Security
English | 简体ä¸æ–‡
🚧 Ordine is currently in Preview. APIs and features may change before beta.
Ordine is an AI Agent first work Orchestration framework that lets you define typed operations, compose them into DAG pipelines, and execute them with any AI agent or script executor.
No more scattered scripts. No more babysitting agent runs. Define your workflow once as a pipeline — then let Claude, GPT, Gemini, or your own agent execute it. Agents are the primary runtime, not an afterthought. Code quality automation ships as a built-in plugin.
- Objects — Typed pipeline inputs (folders, code files, GitHub projects, or custom types via plugins)
- Operations — Atomic tasks with configurable AI agent or script executors
- Pipelines — Chain operations into multi-step DAG workflows
- Skills — Pluggable AI agent capabilities that power operation execution
- Agent — Choose any AI agent as executor — Claude, GPT, Gemini, or your own
- Jobs — Track background execution with real-time progress and traces
- Plugins — Extend with new object types, operations, and domain-specific workflows
```sh bun install cp apps/app/.env.example apps/app/.env cp apps/server/.env.example apps/server/.env
cd apps/app && bun run db:push && cd ../..
bun dev ```
| Service | URL |
|---|---|
| Main app | http://localhost:9430 |
| API server | http://localhost:9433 |
Ordine is in preview. APIs, data models, and workflows may change before beta.
External contributions are paused until the beta release. See CONTRIBUTING.md for the current policy.
Ordine does not have a public security intake process yet. See SECURITY.md for the current preview-stage policy.
Visit the documentation site for guides, API reference, and skill library.
MIT © 2026 Code Forge AI