Frequently Asked Questions
General
What is DaemonCore?
DaemonCore is a deterministic foundation for multi-agent intelligence. It provides the structural layer that enables multiple AI agents to operate with precision, coordination, and safety.
Is DaemonCore an AI model?
No. DaemonCore is not an AI model or assistant. It’s an orchestration and governance layer that AI models (like Claude, GPT, Gemini) operate within. Think of it as the operating system for multi-agent workflows.
What problems does DaemonCore solve?
- Coordination - Multiple agents working together without collision
- Determinism - Predictable, reproducible behaviour
- Safety - Configurable autonomy with approval gates
- Vendor isolation - Clean separation between different AI providers
Technical
What’s a Kernel in DaemonCore?
The kernel is the core specification layer. It defines:
- What agents are allowed to do (capability ceilings)
- How agents must behave (safety invariants)
- How agents identify themselves (identity framework)
- How agents coordinate (exchange bus protocol)
The kernel is declarative configuration, not executable code.
How does boot sequence work?
Agents follow a strict 9-phase boot sequence:
- Kernel load
- Identity handshake
- Environment reporting
- Profile application
- Global context load
- Vendor persona load
- Project boot
- Session resume
- Modulation (optional)
Each phase must complete before the next begins.
What’s the difference between Sandbox, Controlled, and Dangerous modes?
- Sandbox - Read-only operations, all writes require approval
- Controlled - Normal operations, destructive actions require approval
- Dangerous - Full autonomy, audit-only (no blocking gates)
Can I use DaemonCore with my own AI models?
Yes. DaemonCore’s vendor abstraction layer supports multiple AI providers. You define vendor-specific persona and init files, and the kernel handles the rest.
Development
Is DaemonCore open source?
The kernel specification is public. The public reference implementation is being prepared for long-term community use and will be published once that work is complete. Licensing and release timing have not yet been decided.
How do I contribute?
A public contribution process is not open yet. It will be introduced alongside the public reference implementation.
Where can I get help?
- Check this FAQ
- Read the Architecture Overview