Orka
Orchestrate multi-agent AI pipelines directly from VS Code.
Orka lets you define fleets of AI agents, run them through gated stages, review outputs at each step, and ship — all without leaving your editor. Build automated SDLC workflows where agents handle planning, implementation, testing, and review in a coordinated pipeline.
Features
- 🚀 Multi-Agent Pipelines — Chain multiple AI agents into sequential or parallel workflows
- 🚧 Gated Stages — Review and approve agent outputs before the pipeline continues
- 💬 Chat Integration — Control everything via
@orka in GitHub Copilot Chat
- 🔧 Fleet Editor — Visual configuration for your agent pipelines
- 📊 Context Flow — Visualize how context passes between agents
- 📋 Sidebar Dashboard — Live progress tracking in the Activity Bar
Requirements
Getting Started
- Install Orka from the VS Code Marketplace
- Open a workspace where you want to run your pipeline
- Type
@orka in Copilot Chat to get started
Quick Start
@orka /start implement user authentication
This kicks off a pipeline with your configured agent fleet. Agents run sequentially, and you approve outputs at each gate before the pipeline continues.
Chat Commands
Type @orka in Copilot Chat followed by a command:
| Command |
Description |
/start <description> |
🚀 Start a new feature pipeline |
/resume |
🔄 Resume a paused pipeline |
/status |
📋 Show status of all features |
/pipeline |
🔮 Show pipeline configuration |
/approve |
✅ Approve the current gate and continue |
/stop |
⏸️ Pause the pipeline |
/feedback <text> |
💬 Send feedback and re-run the current agent |
/fleet |
🔧 Open the Fleet Editor |
Command Palette
Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) and search for:
- Orka: Start Feature — kick off a new pipeline
- Orka: Resume Feature — resume a paused run
- Orka: Show Status — view pipeline status
- Orka: Show Pipeline — inspect pipeline config
- Orka: Edit Fleet — configure your agent fleet
- Orka: Show Context Flow — visualize context passing between agents
Fleet Configuration
Orka ships with a default pipeline, but you can customize it per-project. Use /fleet to open the Fleet Editor, or create a .orka/fleet_config.json in your workspace root.
A fleet config defines:
- Fleets — groups of agents that run as a stage
- Agents — individual AI agents with specific roles
- Gates — approval checkpoints between agents
How It Works
- Start — Describe a feature and Orka creates a pipeline run
- Execute — Agents run one by one (or in parallel), each producing output
- Gate — After each agent, you review the output and choose to approve, provide feedback, or stop
- Context Flow — Each agent's output becomes the next agent's input automatically
- Complete — Once all agents finish, your feature is ready
License
MIT — see LICENSE for details.
