clawdie-ai/README.md

🦞 Clawdie

Personal AI Assistant on FreeBSD

A lean, secure AI assistant running on FreeBSD 15 with native jail isolation.

Documentation  •   Website  •   Codeberg

 

38.4k tokens · 19% of context window   •   built-in knowledge · preloaded

---

Overview

Clawdie is a FreeBSD-first personal AI assistant with a host-side Node.js orchestrator and jailed agent execution. The core repo stays small, Telegram is built in, and additional channels or integrations are added through skills instead of being hard-wired into core.

V2 Track

The main branch stays focused on the current working system. The multitenant branch is the internal V2 track where Clawdie is refactored into a host platform that can run tenant assistants such as atlas or bob.

That branch is for architecture work, not a feature race. The goal is to make ownership, recovery, degraded mode, and FreeBSD-native multitenancy explicit before any wider public launch.

Current Release

v0.10.0 (04.apr.2026) Domain Defaults Fix

See Changelog for full release notes.

Key Features:

• 🏠 Native FreeBSD - No Linux emulation, native jail isolation
• 🔒 Jail Isolation - Secure OS-level containers with ZFS snapshots
• ⚡ Better Performance - Single layer: FreeBSD → Jail → Node.js
• 💾 ZFS Integration - Snapshots, quotas, compression, RAIDZ1 support
• 🌐 Multi-Language Docs - English, German, French, Spanish (Crowdin-managed)
• 📚 Zero-Downtime Deployments - Symlink-based versioning with instant rollback
• 🔄 Automated Sync - Daily documentation compilation & deployment @ 05:00 UTC
• 🔧 Cloud & Baremetal - ISO variants for VPS (headless) and physical hardware (GUI)
• 🔐 SSH Keys & Passwords - Pre-configured or interactive setup options
• 🔧 CNC Ready - clawdie-cnc for CNC machine control (in development)

Why Clawdie?

Why Jails Instead of Docker?

Aspect	Docker on FreeBSD	FreeBSD Jails
Performance	Good (emulation overhead)	Excellent (native)
Layers	FreeBSD → Linux → Docker	FreeBSD → Jail
Security	Good	Excellent
Networking	Docker bridge	Native IP
Complexity	High	Low
ZFS Integration	Manual	Native

Result: Simpler, faster, more secure.

Why Clawdie?

NanoClaw by Gavriel is the upstream line we still track, and NanoClaw itself follows the broader OpenClaw direction opened by Peter Steinberger. Clawdie takes that line onto FreeBSD for operators who want:

• Native performance without Linux emulation
• ZFS integration for snapshots and quotas
• Simpler architecture (fewer layers)
• CNC machine integration for manufacturing

Architecture

Telegram + scheduled tasks + IPC
  ↓
PostgreSQL ops database (messages, groups, tasks, sessions)
  ↓
Host orchestrator (Node.js)
  ↓
FreeBSD worker jail(s)
  ↓
Agent response

Single Node.js process. Agents execute in isolated FreeBSD worker jails with filesystem isolation via nullfs mounts. Only mounted directories are accessible. Per-group message queue with concurrency control. PostgreSQL defaults to host runtime (DB_RUNTIME=host) with platform-owned system_* databases: built-in skills/reference knowledge, user/agent memory, and operational state. A dedicated db jail remains available when explicitly selected. Ops + memory are required; the built-in skills artifact is committed and can be refreshed with just refresh-skills-artifact when docs or skills change.

Key Files

• src/index.ts — orchestrator: state, message loop, agent invocation
• src/hostd/ — privileged host daemon (types, ops, daemon, client)
• src/controlplane.ts — self-healing: checks hostd, service jails, PF; repairs via hostd
• src/watchdog.ts — memory throttle, run mode, control plane timer
• src/channels/telegram.ts — built-in Telegram channel
• src/group-queue.ts — per-group queue with global concurrency limit
• src/agent-runner.ts — spawns jailed agent processes
• src/mount-security.ts — mount allowlist enforcement
• src/task-scheduler.ts — scheduled tasks
• src/db.ts — PostgreSQL ops (messages, groups, sessions, state, routing)
• setup/*.ts — modular setup entrypoints
• setup/service.ts — generates run-*.sh wrappers at install time (not tracked in git)
• groups/*/AGENTS.md — per-group memory

A justfile provides discoverable command shortcuts. Run just (or just --list) to see all recipes. Prefer just for interactive work; npm run <name> remains for CI and scripting. The full installer is intentionally exposed as just install / npm run install:clawdie instead of the npm lifecycle name install.

Common mappings:

just install              # npm run install:clawdie
just install-from-db      # npm run install:clawdie -- --from db
just setup-db             # npm run setup -- --step db
just setup-cms            # npm run setup -- --step cms

What It Supports

• Telegram-first core - Built-in Telegram channel plus headless/background operation.
• Isolated group context - Each group has its own AGENTS.md memory, isolated filesystem, and runs in its own jail sandbox with only that filesystem mounted.
• Main channel - Your private channel (self-chat) for admin control; every group is completely isolated.
• Scheduled tasks - Recurring jobs that run the AI agent and can message you back.
• Jail isolation - Agents are sandboxed in FreeBSD jails with ZFS snapshots.
• Layered Memory Fabric - PostgreSQL stores three platform-owned planes by default: system_skills for reviewed skills/reference knowledge, system_brain for user/agent memory, and system_ops for messages, tasks, sessions, routing, and runtime state. See docs/public/architecture/layered-memory-fabric.md.
• Built-in payments, optional extras - Stripe ships in core and is ready when configured; other channels and integrations can still be layered through skills.

Installation

Prerequisites

• FreeBSD 15.x
• ZFS root installation
• Root access or sudo
• Internet access for pkg and npm
• bash and git available before first run
• just — command runner (pkg install just on FreeBSD; preinstalled on Clawdie ISO)
• At least one agent CLI on PATH: one of pi, aider, claude, codex, or gemini. The controlplane harness uses Aider+Pi as the primary driver. Onboarding fails fast if none are present (see doc/AGENT-CLI-VALIDATION.md for the validated install paths). The Clawdie ISO ships claude/gemini/pi via the npm-globals bundle, aider via py311-aider_chat pkg, and codex via pkg install codex.

Recommended explicit host baseline before first run:

sudo pkg install -y bash git bastille node24 npm tmux btop python311 uv ripgrep fd-find rsync postgresql18-client py311-pillow dejavu py311-aider_chat edk2-bhyve just

The edk2-bhyve package provides UEFI firmware required for optional browser-vm support (FEATURE_BHYVE_GUI=YES in .env).

On FreeBSD, use fd-find. It provides the fd command that pi expects and avoids colliding with the unrelated fd file manager package.

The host baseline also includes py311-pillow and dejavu so tmux screenshot capture works without a separate uv pip install Pillow step.

If the host still has both Python 3.11 and 3.12 installed, pin uv to 3.11 explicitly until the generic python3 path is cleaned up:

uv venv --python 3.11
uv run --python 3.11 <command>

Operator Glasspane

The local operator cockpit lives in scripts/glass.sh. On a root/platform install it uses the tmux session name glasspane. On a tenant install it uses <tenant>-glasspane.

# create and attach
sh scripts/glass.sh

# create without attaching (safe smoke test)
GLASS_ATTACH=0 sh scripts/glass.sh

# remove the glasspane session
sh scripts/glass.sh kill

On FreeBSD, keep fd-find installed. It provides the fd command that pi expects inside the glasspane session.

First use:

• Start it with sh scripts/glass.sh.
• Window 0 (glass) contains pi on the left and aider on the right.
• Window 1 contains btop.
• Switch windows with Ctrl-b 0 and Ctrl-b 1.
• Move between panes with Ctrl-b then an arrow key.
• Detach without killing the session with Ctrl-b d.
• Reattach later with tmux attach -t glasspane.
• Stop the session with sh scripts/glass.sh kill.

Quick Start

# 1. Install the recommended FreeBSD host baseline
sudo pkg install -y bash git bastille node24 npm tmux btop python311 uv ripgrep fd-find rsync postgresql18-client py311-pillow dejavu py311-aider_chat just

# 2. Clone the repository
git clone https://codeberg.org/Clawdie/Clawdie-AI.git /home/clawdie/clawdie-ai
cd /home/clawdie/clawdie-ai

# 3. Bootstrap repo dependencies and hooks
./setup.sh

# 4. Ensure at least one agent CLI is on PATH (onboarding gates on this).
#    Pick whichever you use; the ISO bundle installs all supported CLIs offline.
pi --version >/dev/null 2>&1 || npm install -g @earendil-works/pi-coding-agent
# Optional alternates:
#   npm install -g @anthropic-ai/claude-code   # claude
#   npm install -g @google/gemini-cli          # gemini
#   sudo pkg install -y codex                  # codex (FreeBSD pkg)

# 5. Run the canonical install orchestrator
just install

# 6. If a step fails, fix the cause and resume from that step
just install-from-db

# 7. Validate the result
just doctor

If the host still has the old fd file manager package installed, pkg will replace it with fd-find. That replacement is the intended FreeBSD baseline.

just install is the supported full path. The lower-level setup/index.ts steps remain available for debugging and targeted recovery, but new installs should prefer the orchestrator so checkpoints and resume behavior stay consistent.

PI Profiles

pi-tui uses a single PI_TUI_PROFILE selector for common runtime scenarios:

• setup — onboarding and exact setup actions
• operator — normal daily runtime
• status — read-only health/status summaries
• payments — Stripe-safe support flow
• docs — docs and release copy
• cms — Astro/Strapi publishing work
• git — local git mirror and upstream sync
• memory — Layered Memory Fabric and PostgreSQL memory work
• local — Ollama/offline fallback
• cnc — CNC machine operation and G-code work

Check the resolved profile state with:

just setup -- --step pi-config

pi-config validates the active provider through Pi auth in ~/.pi/agent/auth.json, provider env vars in .env, or OLLAMA_HOST for local Ollama.

List all built-in profiles with:

just setup -- --step pi-config --list-profiles

The Layered Memory Fabric database path is mandatory. Setup generates or preserves the PostgreSQL and Strapi secrets in .env, then keeps SKILLS_DB_URL, MEMORY_DB_URL, and OPS_DB_URL aligned with AGENT_NAME, subnet, and passwords. The db step applies three distinct schema tracks:

• built-in knowledge tables in the skills DB
• dynamic hybrid-memory tables in the user/agent memory DB
• operational state tables in the ops DB (messages, tasks, sessions, routing)

The db step also imports the shipped built-in knowledge artifact by default. skills-memory --import remains available as a manual re-import path. Maintainers refresh the committed artifact with just refresh-skills-artifact; the wrapper detects docs/skills/identity changes and checks OpenRouter budget before regenerating embeddings. At runtime, the host queries the local skills DB before each jail run and prepends the top built-in references to the prompt when relevant.

The active mandatory pieces are:

• POSTGRES_ADMIN_PASSWORD
• SKILLS_DB_PASSWORD
• MEMORY_DB_PASSWORD
• OPS_DB_PASSWORD
• CONTROLPLANE_SHARED_SECRET

For PostgreSQL identifiers, the root install uses the shared platform namespace system, independent of ASSISTANT_NAME or AGENT_NAME. Platform databases use system_skills, system_brain, system_ops, system_git, and system_web. Later additive tenants get their own <tenant>_* databases.

STRAPI_DB_PASSWORD and the STRAPI_* app secrets feed the default internal Strapi bootstrap inside the cms jail. They are generated automatically if missing and should stay private to the CMS layer.

npm run preflight-check writes a full host validation bundle into tmp/preflight/<stamp>/, including summary.json and summary.env for the on-host agent. It includes telegram-auth in the default sequence; groups and register remain post-deployment follow-up work and no longer block base install verification on a fresh host.

Clawdie provisions local code hosting by default through the dedicated git jail. The default mode is:

• CODE_HOSTING_MODE=git
• FEATURE_GIT=YES
• WARDEN_GIT_IP=<subnet>.2

The git step creates the shared git jail by default (GIT_JAIL_NAME can override it), installs the git jail baseline, creates /srv/git, and mirrors the current repository into a bare repository.

Default IP Layout

The repo registry/default examples use 10.0.1.x; live installs can override this through .env with another private /24 such as 192.168.72.x. For a running system, live jail state and .env are authoritative; this table is the provisioning/default slot map from infra/jails.yaml:

• .1 gateway on warden0
• .2 git / Git Service
• .3 cms / Web Service
• .4 Local AI Models (ollama or llama.cpp, if enabled)
• .5 optional db / Data Service when DB_RUNTIME=jail
• .6 browser / Browser Execution Template
• .101+ workers
• .211 DB admin worker
• .212 Git admin worker
• .213 coordinator worker
• .150+ ephemeral browser task clones

This keeps foundational services low in the range while leaving room for multiple workers and avoiding overlap with host-facing web services. The CMS stack stays on the private jail network instead of assuming control of host nginx, which is the main reason the low fixed service slots matter.

Naming and Local DNS

Current main separates public and internal naming:

• AGENT_DOMAIN is the public-facing site/API domain
• AGENT_INTERNAL_DOMAIN is the internal jail/service zone written into /etc/hosts

Fresh installs should keep public and internal names separate:

• AGENT_DOMAIN= until you have a real public DNS name configured
• AGENT_INTERNAL_DOMAIN=<agent>.home.arpa for local jail/service names

Why this changed:

• .local is reserved for mDNS and is a bad default for private service naming
• home.arpa is the safer standards-based internal namespace
• keeping public and internal names separate avoids accidentally treating a non-public jail hostname as if it were externally routable

FreeBSD Milestone

Example tmux screenshot generated by Clawdie and sent as a Telegram photo

Current runtime centers on:

• host-side orchestration in src/index.ts
• service-owned shared infrastructure (clawdie service, system_* platform resources)
• Bastille worker jails from src/jail-config.ts
• just install as the canonical install orchestrator
• mandatory DB bootstrap with built-in knowledge import when the artifact is present

Simple Install Flow

./setup.sh
  -> bootstrap repo dependencies and hooks
just install
  -> read first-boot setup seed or run onboarding fallback
  -> configure environment, pi, PF, jails, services, hostd
  -> provision Data Service, Git Service, Web Service when enabled
  -> verify install and report warnings

ASSISTANT_NAME is display-only. The root install remains the shared platform: service name clawdie, platform namespace system, and no root TENANT_ID. groups and register remain explicit follow-up steps after auth/state are present; they are not part of the minimal first-boot path.

Stripe Flow

onboarding
  -> configure Stripe now? (or skip)
  -> write STRIPE_SECRET_KEY into .env
host agent-runner
  -> reads STRIPE_SECRET_KEY from .env
  -> passes it to jailed runtime in stdin JSON secrets payload
jailed agent runner
  -> merges secret into SDK env only
  -> registers Stripe MCP tools
agent chat
  -> can use payment, customer, invoice, and subscription tools

Web Serving

nginx runs inside the Web Service (cms jail). By default, setup/pf.ts configures Warden egress (NAT) only; public ingress (PF rdr) is intentionally opt-in so Clawdie does not take over host ports 80/443 unexpectedly. Strapi (when used) is internal-only by default.

Operational docs for this milestone:

• Bastille on FreeBSD 15
• Jail Networking Strategy
• PostgreSQL Memory Plan
• Monitoring Model
• Manual Reset and pi-tui Rebootstrap

Usage

Talk to your assistant with the trigger word (default: @Clawdie):

@Clawdie send an overview of the sales pipeline every weekday morning at 9am
@Clawdie review the git history for the past week each Friday and update the README
@Clawdie every Monday at 8am, compile news on AI developments and message me a briefing

Run just doctor to check system health (runtime, jails, networking, services, DNS resolver probes, TLS certificate expiry, ACME renewal cron, morning-report schedule, and Layered Memory Fabric DB/artifact status). Exit codes: 0 = ok or warnings, 1 = error.

TLS Certificate Lifecycle

Public edge certificates are managed by acme.sh on the FreeBSD host and installed under /usr/local/etc/nginx/ssl/<label>/. The TLS setup step is safe by default:

# Non-mutating local validation: acme binary, cron, nginx -t, webroots, challenge routes, cert parsing
sudo npx tsx setup/index.ts --step tls --smoke-test

# Repair/install lifecycle without forcing issuance: install acme.sh, install cert hooks, write /etc/cron.d/clawdie-acme
sudo npx tsx setup/index.ts --step tls --apply

# Force-renew one cert only when deliberately validating or recovering before expiry
sudo npx tsx setup/index.ts --step tls --apply --cert clawdie --force-renew

just doctor reports TLS_<LABEL> days remaining and ACME_RENEWAL_CRON status. Do not force-renew repeatedly; Let’s Encrypt rate-limits duplicate certificates.

From the main channel (your self-chat), you can manage groups and tasks:

@Clawdie list all scheduled tasks across groups
@Clawdie pause the Monday briefing task
@Clawdie join the Family Chat group

Telegram Commands

A short selection — for the full reference (status, structured reports, runtime, sessions, admin actions, free-text routing) see Operator Commands.

Command	Description	Auth
/status	System summary: jails, ZFS, PF, budget, model	anyone
/report	Structured system + auth report	admin
/disk	Structured ZFS pool + snapshot report	admin
/tasks	Structured controlplane task report	admin
/budgetreport	Structured budget + token analytics	admin
/publishreport	Structured tenant publish/content report	admin
/testreport	Structured build + test status (from wrapper-written JSON)	admin
/policy	Default runtime, per-chat overrides, fallback cooldowns	anyone
/usage	Per-agent token budget breakdown	anyone
/clearcooldown	Clear a provider fallback cooldown	ops chat
/compact	Compact session (summarize old, keep recent turns)	admin
/new	Hard reset session, start fresh	admin
/stop	Kill running agent mid-response	admin
/tts	Toggle voice replies (on/off/status/default)	admin
/activation	Set trigger mode (always/mention)	admin
/whoami	Show your Telegram identity	anyone
/help	List available commands	anyone

The bot also routes free-text ops phrasings ("disk usage", "are the tests passing", "task report", etc.) to the matching structured report instead of the LLM path — see Structured Reports → Free-Text Routing.

Session Compaction

When a conversation grows past the byte limit (AGENT_SESSION_MAX_BYTES), old turns are automatically summarized via the LLM and stored in the memory database. The session file is rewritten with a compaction header + last N turns at full fidelity. Manual compaction is available via /compact.

Cost Model

Budget enforcement has been removed (26.maj.2026). DeepSeek prefix caching at ~98% hit rate makes per-chat token budgets unnecessary.

Text-to-Speech

Clawdie uses Microsoft Edge TTS (free, no API key) to generate voice replies. Configured via TTS_AUTO_MODE:

• always — every reply gets a voice message
• inbound — reply with voice when user sends a voice note
• tagged — only when agent output contains [[tts]] marker
• off — no voice messages

Per-chat override via /tts command.

Customizing

Clawdie doesn't use configuration files. To make changes, just tell the AI agent what you want:

• "Change the trigger word to @Assistant"
• "Remember in the future to make responses shorter and more direct"
• "Add a custom greeting when I say good morning"
• "Store conversation summaries weekly"

Or run /customize for guided changes.

The codebase is small enough that the AI agent can safely modify it.

Philosophy

Small enough to understand. One process, a few source files, no microservices. If you want to understand the full Clawdie codebase, ask the AI agent to walk you through it.

Secure by isolation. Agents run in FreeBSD jails with filesystem isolation via nullfs mounts. They can only access explicitly mounted directories. Bash access is safe because commands run inside the jail, not on your host.

Built for the individual user. Clawdie isn't a monolithic framework; it's software that fits each user's exact needs. You make your own fork and have the AI agent modify it to match your requirements.

Customization = code changes. No configuration sprawl. Want different behavior? Modify the code. The codebase is small enough that it's safe to make changes.

AI-native.

• FreeBSD installs can start with npm run wizard for optional interactive locale, timezone, and assistant-first onboarding. bsddialog is used when present; plain TTY prompts are the fallback. The selector uses FreeBSD's locale inventory, so you can choose any installed locale (for example en_US.UTF-8 or zh_CN.UTF-8) and it is applied the same way as the default.
• The rest of setup remains modular and scriptable through just setup -- --step ... (or npm run setup -- --step ...).
• No monitoring dashboard; the harness shows status and the AI fills in the details.
• No debugging tools; describe the problem and the AI fixes it.

Skills over features. Instead of adding features (e.g., support for Telegram) to the codebase, contributors submit skills like /add-telegram that transform your fork. You end up with clean code that does exactly what you need.

Documentation System (Phase 3.0+)

v0.9.0 includes a complete multi-language documentation deployment system:

Architecture

Markdown Source (git)    →    Crowdin (translate)    →    Daily Sync (05:00 UTC)
     ↓                              ↓                              ↓
docs/public/*.md (English)    →    docs/public/{de,fr,es}/* →    docs-v0.9.0_24.mar.2026/
                                                          ├─ en/
                                                          ├─ de/
                                                          ├─ fr/
                                                          └─ es/

Symlink Swap (atomic, zero-downtime)
     ↓
https://docs.clawdie.si/{en,de,fr,es}/

Features

• Markdown Source of Truth — English docs in docs/public/*.md, auto-synced to Crowdin
• Crowdin Integration — Translators work on crowdin.com/project/clawdie-ai
• Automated Sync — Daily cron @ 05:00 UTC compiles markdown → HTML for all languages
• Zero-Downtime Deployment — Symlink-based versioning with atomic swaps
• 30-Day Retention — Old versions auto-cleaned, instant rollback available
• No External Dependencies — Pure shell compilation (no pandoc, no external tools)

Quick Links

• docs-deployment skill — Complete operations guide
• CROWDIN.md — Translation workflow
• Integration Guide — How build (clawdie-shell) and deployment (clawdie-ai) work together

Supported Languages

• 🇬🇧 English (source, always current)
• 🇩🇪 Deutsch (German)
• 🇫🇷 Français (French)
• 🇪🇸 Español (Spanish)

Want to add more languages? Update scripts/docs-sync.cron.sh and add to Crowdin.

---

Clawdie Ecosystem

Component	Description	Status
{AGENT_NAME}-worker	Default Bastille worker jail profile	Current
skills-memory bootstrap	Precomputed pgvector import for the skills DB	Current
docs-deployment	Multi-language docs with Crowdin sync (Phase 3.0+)	Live
clawdie-cnc	CNC machine control	Development
browser-vm	Browser automation profile in jail config	Planned
docs.clawdie.si	Documentation site (4 languages)	Live

CNC (clawdie-cnc)

Clawdie-cnc is the control system for the CNC machine — an open source machine designed to manufacture geodesic dome components. It translates dome designs from OSA into G-code, manages the build sequence, and logs everything to persistent memory.

See clawdie-cnc/README.md for details.

Contributing

Don't add features. Add skills.

If you want to add Slack support, don't create a PR that adds Slack alongside Telegram. Instead, contribute a skill file (.agent/skills/add-slack/SKILL.md) that teaches the AI agent how to transform a Clawdie installation to use Slack.

Users then run /add-slack on their fork and get clean code that does exactly what they need, not a bloated system trying to support every use case.

Available Skills (35+)

Channels: add-telegram, add-discord, add-gmail, add-slack, add-voice-transcription, x-integration

Jail & Infrastructure: warden-bootstrap, warden-pf, warden-zfs, warden-health, bastille-network, browser-vm, freebsd-admin, sanoid

Operations:

• nginx — CMS jail vhost management
• nginx-glasspane — Nginx security features
• docs-deployment — Multi-language documentation sync (NEW v0.9.0)
• telegram-admin, tmux-screenshot, postgres-memory

Frontend & CMS: astro, strapi

Agent: coding-agent, add-telegram-swarm, add-parallel

Utility: setup, update, customize, debug

Skills Structure

.agent/skills/add-telegram/
├── SKILL.md           # Skill documentation and instructions
├── manifest.yaml      # Skill metadata
├── add/               # Files to add
│   └── src/channels/telegram.ts
├── modify/            # Files to modify
│   └── src/config.ts
└── tests/             # Skill tests
    └── telegram.test.ts

Requirements

• FreeBSD 15.x
• Node.js 24+
• Python 3.11+ (for voice transcription, screenshots, and other features)
• ZFS (required for Bastille jails)
• Optional: Telegram bot token, provider API key, PostgreSQL URLs for memory/skills DB

Security

Agents run in FreeBSD jails with filesystem isolation, not behind application-level permission checks. They can only access explicitly mounted directories. The codebase is small enough that you can review it. See docs/public/operate/security.md for the full security model.

Documentation

• docs.clawdie.si - Complete documentation (English, Deutsch, Français, Español)
• docs/public/architecture/host-operator-model.md - Host-first operator model and service-jail layout
• .agent/skills/docs-deployment/SKILL.md - Documentation deployment system (Phase 3.0+)
• .agent/skills/docs-deployment/INTEGRATION.md - Cross-repo integration guide (clawdie-ai ↔ clawdie-shell)
• CROWDIN.md - How to translate and contribute translations
• AGENTS.md - Agent development guidelines

Community

Questions or ideas? Open an issue on Codeberg or reach out at hello@clawdie.si.

Acknowledgments

Clawdie is built on giants' shoulders. Peter Steinberger created OpenClaw, which set the broader project line. NanoClaw by Gavriel distilled that line into a minimal personal-assistant upstream. Clawdie is the FreeBSD-first fork in that lineage.

We keep NanoClaw as an upstream reference where it helps, then carry the design into native jails, ZFS, PF, and the wider OSA mission.

License

BSD 3-Clause License

Copyright (c) 2026, Sam (Samo Blatnik)

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

1. Redistributions of source code must retain the above copyright notice, this list of conditions, and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions, and the following disclaimer in the documentation and/or other materials provided with the distribution.
3. Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

See LICENSE for full text.

---

Clawdie · Personal AI Assistant on FreeBSD
clawdie.si · docs.clawdie.si
Built on giants' shoulders · FreeBSD-first