How It Works
For Developers (Building & Deploying)
For Teams (Running Agents)
The Four Types
orchagent has four canonical types:
Simple rule:
- Want a single LLM answer? Use
type: "prompt"(orch run) - Want your code to execute? Use
type: "tool"(orch run) - Want the LLM to iterate with tools? Use
type: "agent"(orch run) - Want to share knowledge? Use
type: "skill"(orch skill install) - Want to DELEGATE to sub-agents? Export agents (
orch install)
- On-demand or always-on (
run_mode) — each call is independent, or deploy as a service - Specific - one task, done well
- Callable - simple API or CLI call; optionally callable by other agents
Key Features
BYOK: Bring Your Own Key
orchagent uses a Bring Your Own Key model. Developers provide their own LLM API keys (OpenAI, Anthropic, Gemini). Why BYOK:- Authors don’t pay for others’ usage
- No middleman markup on LLM costs
- Direct billing with your LLM provider
API keys are separate from subscriptions (Claude Pro/Max, ChatGPT Plus). You’ll need to set up API billing at console.anthropic.com or platform.openai.com.
Next Steps
Quickstart
Deploy your first agent in minutes
API Reference
Authentication, endpoints, errors
CLI Commands
orch login, run, publish
Agent Types
The four types — prompt, tool, agent, skill
orch-hq
Desktop command centre
Scheduling
Cron schedules and webhooks