FreeBSD-native AI assistant
Self-hosted AI on FreeBSD with split-brain memory and native jail isolation
Clawdie is a self-hosted AI system built for FreeBSD. It follows the NanoClaw upstream line, but runs with native jails, PF, ZFS, and a local built-in knowledge layer designed to reduce first-install LLM setup friction.
Local-first AI operations on FreeBSD, not another Linux-only stack.
Prebuilt built-in knowledge in the database jail before production LLM keys are added.
Upstream-aware through NanoClaw, but not upstream-dependent in deployment.
Split brain is simple: Clawdie treats built-in system knowledge and user memory as two different jobs.
This is the memory Clawdie ships with: install help, system docs,
operator workflows, and product skills. It is prepared ahead of
time and imported into the db jail during bootstrap.
This is the changing side: user preferences, notes, later personal
agents like mevy or bob, and other runtime
memories that grow after install.
Clawdie arrives with her own instruction book already packed. Later, she learns your household habits separately, instead of mixing them into the factory manual.
Brain A is a prebuilt vector-backed knowledge package stored locally in the database jail. Brain B is dynamic user and agent memory with a separate lifecycle, provenance, and update path.
Most AI tooling assumes Linux by default. Clawdie uses FreeBSD on purpose: jails, PF, and ZFS are not wrappers here. They are core operating-system features. That makes Clawdie a serious alternative for operators who want platform diversity and tighter system control.
Large infrastructure operators often avoid monoculture because shared bugs create shared outages. Clawdie follows the same logic for self-hosted AI: stay close to the Linux-born upstream ecosystem, but provide a FreeBSD deployment path with a different failure surface.
Isolation without a Linux container stack layered on top.
Rollback points for installs, experiments, and service changes.
Network control as part of the system design, not an afterthought.
Operationally attractive for long-lived infrastructure work.
Clawdie did not appear from nowhere. It grows from NanoClaw and keeps track of upstream changes. But the goal is not to become a thin mirror. The goal is to evolve a FreeBSD-native operating path while still learning from upstream improvements.
| Topic | NanoClaw upstream | Clawdie |
|---|---|---|
| Primary ecosystem | Linux-first | FreeBSD-first deployment |
| Upstream tracking | Source of evolution | Fetch-only review, operator decides |
| Onboarding goal | General framework | Lower-friction bootstrap with local built-in knowledge |
| Memory model | Generic | Split brain: built-in knowledge plus user memory |
Clawdie is not only a chat interface. The long-term direction is AI that can support real operator work: system setup, maintenance, documentation lookup, machine procedures, and later CNC-related task guidance. Split brain matters here because stable operating knowledge should not be mixed blindly with changing personal memory.
Brain A can carry the stable operating playbook. Brain B can carry shop, operator, or agent-specific context. That separation makes the system easier to trust, audit, and grow.
Architecture, quick reference table, and the two-server setup explained.
ArchitecturePlain-language and technical explanation of Brain A and Brain B.
Core conceptSetup wizard walkthrough, jail provisioning, IP scheme, and screenshots.
Getting startedPlatform setup and host-side FreeBSD operational guidance.
FreeBSD