- Shared assistant brain across channels
- Retrieval-augmented generation (RAG) and document intelligence
- Persistent semantic and typed workspace memory
- Sandboxed tool execution (bash, browser_*, file ops)
- Scheduler and proactive jobs (cron/at/every)
- Persona and skills engine (SOUL.md, SKILL.md)
- Local model backend support (Ollama, LM Studio, vLLM)
- Tamper-evident audit trail and observability exports