The standard for multi-agent systems. Define agents, coordinate workflows, and build systems where agents work together.
⚠️ Experimental: This project is still in active development. Interfaces and APIs may change without notice.
Documentation | Architecture | npm
The OS Protocol is an open-source specification for orchestrating, managing, and executing AI agents in distributed environments. It defines standardized interfaces across six domains:
- System — registry, environment, filesystem, sandbox, settings, preferences, installer, MCP client
- Context — read-only facades (system context, embeddings, key-value)
- Actions — write facades (system actions, tools, MCP servers)
- Checks — quality assurance (rules, judge, audit, screenshot)
- Workflows — execution patterns (routing, parallelization, orchestrator-workers, evaluator-optimizer)
- Runs — lifecycle control (timeout, retry, cancel, approval)
This project includes skills for local development with Claude Code:
| Skill | Description |
|---|---|
/ask |
Answer questions about the protocol — architecture, concepts, rationale, interfaces, domains, and design principles. |
/docs |
Write, edit, review, and triage documentation pages for the website. |
Skills are defined in .claude/skills/ and provide domain-specific context so the agent can work accurately with the protocol.
The documentation site exposes machine-readable endpoints for LLM tools and agents:
| Endpoint | Description |
|---|---|
/llms.txt |
Documentation index — page titles, descriptions, and links grouped by section. |
/llms-full.txt |
Full documentation content — complete markdown text of every page. |
Maintained by @synerops and open to contributions from the community.