CloudClaw documentation

CloudClaw is the operating layer for building, connecting, and governing AI agents across tools, applications, and customer workflows.

Start with one working agent#

1. Obtain a tenant API key through the approved onboarding flow.
2. Create one narrowly scoped agent.
3. Grant only the tools required for its first workflow.
4. Test with synthetic or approved data.
5. Expand access only after the result and audit evidence pass.

CloudClaw follows least-privilege defaults. Start narrow, validate behavior, then add capabilities deliberately.

Core capabilities#

• Agents with focused instructions and explicit tools.
• Connectors with tenant and permission boundaries.
• MCP interoperability for compatible clients.
• Webhooks for event-driven automation.
• Structured errors and request IDs for support.