ZeroClaw 🦀

Zero overhead. Zero compromise. 100% Rust. 100% Agnostic.
⚡️ Runs on $10 hardware with <5MB RAM: That's 99% less memory than OpenClaw and 98% cheaper than a Mac mini!

Built by students and members of the Harvard, MIT, and Sundai.Club communities.

🌐 Languages: 🇺🇸 English · 🇨🇳 简体中文 · 🇯🇵 日本語 · 🇰🇷 한국어 · 🇻🇳 Tiếng Việt · 🇵🇭 Tagalog · 🇪🇸 Español · 🇧🇷 Português · 🇮🇹 Italiano · 🇩🇪 Deutsch · 🇫🇷 Français · 🇸🇦 العربية · 🇮🇳 हिन्दी · 🇷🇺 Русский · 🇧🇩 বাংলা · 🇮🇱 עברית · 🇵🇱 Polski · 🇨🇿 Čeština · 🇳🇱 Nederlands · 🇹🇷 Türkçe · 🇺🇦 Українська · 🇮🇩 Bahasa Indonesia · 🇹🇭 ไทย · 🇵🇰 اردو · 🇷🇴 Română · 🇸🇪 Svenska · 🇬🇷 Ελληνικά · 🇭🇺 Magyar · 🇫🇮 Suomi · 🇩🇰 Dansk · 🇳🇴 Norsk

Getting Started | One-Click Setup | Docs Hub | Docs TOC

Quick Routes: Reference · Operations · Troubleshoot · Security · Hardware · Contribute

Fast, small, and fully autonomous AI assistant infrastructure
Deploy anywhere. Swap anything.

ZeroClaw is the runtime operating system for agentic workflows — infrastructure that abstracts models, tools, memory, and execution so agents can be built once and run anywhere.

Trait-driven architecture · secure-by-default runtime · provider/channel/tool swappable · pluggable everything

📢 Announcements

Use this board for important notices (breaking changes, security advisories, maintenance windows, and release blockers).

✨ Features

• 🏎️ Lean Runtime by Default: Common CLI and status workflows run in a few-megabyte memory envelope on release builds.
• 💰 Cost-Efficient Deployment: Designed for low-cost boards and small cloud instances without heavyweight runtime dependencies.
• ⚡ Fast Cold Starts: Single-binary Rust runtime keeps command and daemon startup near-instant for daily operations.
• 🌍 Portable Architecture: One binary-first workflow across ARM, x86, and RISC-V with swappable providers/channels/tools.

Why teams pick ZeroClaw

• Lean by default: small Rust binary, fast startup, low memory footprint.
• Secure by design: pairing, strict sandboxing, explicit allowlists, workspace scoping.
• Fully swappable: core systems are traits (providers, channels, tools, memory, tunnels).
• No lock-in: OpenAI-compatible provider support + pluggable custom endpoints.

Benchmark Snapshot (ZeroClaw vs OpenClaw, Reproducible)

Local machine quick benchmark (macOS arm64, Feb 2026) normalized for 0.8GHz edge hardware.

Notes: ZeroClaw results are measured on release builds using /usr/bin/time -l. OpenClaw requires Node.js runtime (typically ~390MB additional memory overhead), while NanoBot requires Python runtime. PicoClaw and ZeroClaw are static binaries. The RAM figures above are runtime memory; build-time compilation requirements are higher.

Reproducible local measurement

Benchmark claims can drift as code and toolchains evolve, so always measure your current build locally:

cargo build --release
ls -lh target/release/zeroclaw

/usr/bin/time -l target/release/zeroclaw --help
/usr/bin/time -l target/release/zeroclaw status

Example sample (macOS arm64, measured on February 18, 2026):

• Release binary size: 8.8MB
• zeroclaw --help: about 0.02s real time, ~3.9MB peak memory footprint
• zeroclaw status: about 0.01s real time, ~4.1MB peak memory footprint

Prerequisites

curl -LsSf https://raw.githubusercontent.com/zeroclaw-labs/zeroclaw/master/install.sh | bash

./install.sh --prefer-prebuilt

./install.sh --prebuilt-only

Quick Start

Homebrew (macOS/Linuxbrew)

brew install zeroclaw

One-click bootstrap

# Recommended: clone then run local bootstrap script
git clone https://github.com/zeroclaw-labs/zeroclaw.git
cd zeroclaw
./install.sh

# Optional: bootstrap dependencies + Rust on fresh machines
./install.sh --install-system-deps --install-rust

# Optional: pre-built binary first (recommended on low-RAM/low-disk hosts)
./install.sh --prefer-prebuilt

# Optional: binary-only install (no source build fallback)
./install.sh --prebuilt-only

# Optional: run onboarding in the same flow
./install.sh --onboard --api-key "sk-..." --provider openrouter [--model "openrouter/auto"]

# Optional: run bootstrap + onboarding fully in Docker-compatible mode
./install.sh --docker

# Optional: force Podman as container CLI
ZEROCLAW_CONTAINER_CLI=podman ./install.sh --docker

# Optional: in --docker mode, skip local image build and use local tag or pull fallback image
./install.sh --docker --skip-build

Remote one-liner (review first in security-sensitive environments):

curl -fsSL https://raw.githubusercontent.com/zeroclaw-labs/zeroclaw/master/install.sh | bash

Details: docs/setup-guides/one-click-bootstrap.md (toolchain mode may request sudo for system packages).

Pre-built binaries

Release assets are published for:

• Linux: x86_64, aarch64, armv7
• macOS: x86_64, aarch64
• Windows: x86_64

Download the latest assets from: https://github.com/zeroclaw-labs/zeroclaw/releases/latest

Example (ARM64 Linux):

curl -fsSLO https://github.com/zeroclaw-labs/zeroclaw/releases/latest/download/zeroclaw-aarch64-unknown-linux-gnu.tar.gz
tar xzf zeroclaw-aarch64-unknown-linux-gnu.tar.gz
install -m 0755 zeroclaw "$HOME/.cargo/bin/zeroclaw"

git clone https://github.com/zeroclaw-labs/zeroclaw.git
cd zeroclaw
cargo build --release --locked
cargo install --path . --force --locked

# Ensure ~/.cargo/bin is in your PATH
export PATH="$HOME/.cargo/bin:$PATH"

# Quick setup (no prompts, optional model specification)
zeroclaw onboard --api-key sk-... --provider openrouter [--model "openrouter/auto"]

# Or interactive wizard
zeroclaw onboard --interactive

# If config.toml already exists and you intentionally want to overwrite it
zeroclaw onboard --force

# Or quickly repair channels/allowlists only
zeroclaw onboard --channels-only

# Chat
zeroclaw agent -m "Hello, ZeroClaw!"

# Interactive mode
zeroclaw agent

# Start the gateway (webhook server)
zeroclaw gateway                # default: 127.0.0.1:42617
zeroclaw gateway --port 0       # random port (security hardened)

# Start full autonomous runtime
zeroclaw daemon

# Check status
zeroclaw status
zeroclaw auth status

# Generate shell completions (stdout only, safe to source directly)
source <(zeroclaw completions bash)
zeroclaw completions zsh > ~/.zfunc/_zeroclaw

# Run system diagnostics
zeroclaw doctor

# Check channel health
zeroclaw channel doctor

# Bind a Telegram identity into allowlist
zeroclaw channel bind-telegram 123456789

# Get integration setup details
zeroclaw integrations info Telegram

# Note: Channels (Telegram, Discord, Slack) require daemon to be running
# zeroclaw daemon

# Manage background service
zeroclaw service install
zeroclaw service status
zeroclaw service restart

# On Alpine (OpenRC): sudo zeroclaw service install

# Migrate memory from OpenClaw (safe preview first)
zeroclaw migrate openclaw --dry-run
zeroclaw migrate openclaw

Dev fallback (no global install): prefix commands with cargo run --release -- (example: cargo run --release -- status).

Subscription Auth (OpenAI Codex / Claude Code)

ZeroClaw now supports subscription-native auth profiles (multi-account, encrypted at rest).

• Store file: ~/.zeroclaw/auth-profiles.json
• Encryption key: ~/.zeroclaw/.secret_key
• Profile id format: <provider>:<profile_name> (example: openai-codex:work)

OpenAI Codex OAuth (ChatGPT subscription):

# Recommended on servers/headless
zeroclaw auth login --provider openai-codex --device-code

# Browser/callback flow with paste fallback
zeroclaw auth login --provider openai-codex --profile default
zeroclaw auth paste-redirect --provider openai-codex --profile default

# Check / refresh / switch profile
zeroclaw auth status
zeroclaw auth refresh --provider openai-codex --profile default
zeroclaw auth use --provider openai-codex --profile work

Claude Code / Anthropic setup-token:

# Paste subscription/setup token (Authorization header mode)
zeroclaw auth paste-token --provider anthropic --profile default --auth-kind authorization

# Alias command
zeroclaw auth setup-token --provider anthropic --profile default

Run the agent with subscription auth:

zeroclaw agent --provider openai-codex -m "hello"
zeroclaw agent --provider openai-codex --auth-profile openai-codex:work -m "hello"

# Anthropic supports both API key and auth token env vars:
# ANTHROPIC_AUTH_TOKEN, ANTHROPIC_OAUTH_TOKEN, ANTHROPIC_API_KEY
zeroclaw agent --provider anthropic -m "hello"

Architecture

Every subsystem is a trait — swap implementations with a config change, zero code changes.

Runtime support (current)

• ✅ Supported today: runtime.kind = "native" or runtime.kind = "docker"
• 🚧 Planned, not implemented yet: WASM / edge runtimes

When an unsupported runtime.kind is configured, ZeroClaw now exits with a clear error instead of silently falling back to native.

Memory System (Full-Stack Search Engine)

All custom, zero external dependencies — no Pinecone, no Elasticsearch, no LangChain:

The agent automatically recalls, saves, and manages memory via tools.

[memory]
backend = "sqlite"             # "sqlite", "lucid", "postgres", "markdown", "none"
auto_save = true
embedding_provider = "none"    # "none", "openai", "custom:https://..."
vector_weight = 0.7
keyword_weight = 0.3

# backend = "none" uses an explicit no-op memory backend (no persistence)

# Optional: storage-provider override for remote memory backends.
# When provider = "postgres", ZeroClaw uses PostgreSQL for memory persistence.
# The db_url key also accepts alias `dbURL` for backward compatibility.
#
# [storage.provider.config]
# provider = "postgres"
# db_url = "postgres://user:password@host:5432/zeroclaw"
# schema = "public"
# table = "memories"
# connect_timeout_secs = 15

# Optional for backend = "sqlite": max seconds to wait when opening the DB (e.g. file locked). Omit or leave unset for no timeout.
# sqlite_open_timeout_secs = 30

# Optional for backend = "lucid"
# ZEROCLAW_LUCID_CMD=/usr/local/bin/lucid            # default: lucid
# ZEROCLAW_LUCID_BUDGET=200                          # default: 200
# ZEROCLAW_LUCID_LOCAL_HIT_THRESHOLD=3               # local hit count to skip external recall
# ZEROCLAW_LUCID_RECALL_TIMEOUT_MS=120               # low-latency budget for lucid context recall
# ZEROCLAW_LUCID_STORE_TIMEOUT_MS=800                # async sync timeout for lucid store
# ZEROCLAW_LUCID_FAILURE_COOLDOWN_MS=15000           # cooldown after lucid failure to avoid repeated slow attempts

Security

ZeroClaw enforces security at every layer — not just the sandbox. It passes all items from the community security checklist.

Security Checklist

Run your own nmap: nmap -p 1-65535 <your-host> — ZeroClaw binds to localhost only, so nothing is exposed unless you explicitly configure a tunnel.

Channel allowlists (deny-by-default)

Inbound sender policy is now consistent:

• Empty allowlist = deny all inbound messages
• "*" = allow all (explicit opt-in)
• Otherwise = exact-match allowlist

This keeps accidental exposure low by default.

Full channel configuration reference: docs/reference/api/channels-reference.md.

Recommended low-friction setup (secure + fast):

• Telegram: allowlist your own @username (without @) and/or your numeric Telegram user ID.
• Discord: allowlist your own Discord user ID.
• Slack: allowlist your own Slack member ID (usually starts with U).
• Mattermost: uses standard API v4. Allowlists use Mattermost user IDs.
• Nostr: allowlist sender public keys (hex or npub). Supports NIP-04 and NIP-17 DMs.
• Use "*" only for temporary open testing.

Telegram operator-approval flow:

1. Keep [channels_config.telegram].allowed_users = [] for deny-by-default startup.
2. Unauthorized users receive a hint with a copyable operator command: zeroclaw channel bind-telegram <IDENTITY>.
3. Operator runs that command locally, then user retries sending a message.

If you need a one-shot manual approval, run:

zeroclaw channel bind-telegram 123456789

If you're not sure which identity to use:

1. Start channels and send one message to your bot.
2. Read the warning log to see the exact sender identity.
3. Add that value to the allowlist and rerun channels-only setup.

If you hit authorization warnings in logs (for example: ignoring message from unauthorized user), rerun channel setup only:

zeroclaw onboard --channels-only

Telegram media replies

Telegram routing now replies to the source chat ID from incoming updates (instead of usernames), which avoids Bad Request: chat not found failures.

For non-text replies, ZeroClaw can send Telegram attachments when the assistant includes markers:

• [IMAGE:<path-or-url>]
• [DOCUMENT:<path-or-url>]
• [VIDEO:<path-or-url>]
• [AUDIO:<path-or-url>]
• [VOICE:<path-or-url>]

Paths can be local files (for example /tmp/screenshot.png) or HTTPS URLs.

WhatsApp Setup

ZeroClaw supports two WhatsApp backends:

• WhatsApp Web mode (QR / pair code, no Meta Business API required)
• WhatsApp Business Cloud API mode (official Meta webhook flow)

WhatsApp Web mode (recommended for personal/self-hosted use)

1. Build with WhatsApp Web support:

   cargo build --features whatsapp-web
   

2. Configure ZeroClaw:

   [channels_config.whatsapp]
   session_path = "~/.zeroclaw/state/whatsapp-web/session.db"
   pair_phone = "+15551234567"   # optional; omit to use QR flow
   pair_code = ""               # optional custom pair code
   allowed_numbers = ["+1234567890"]  # E.164 format, or ["*"] for all
   

3. Start channels/daemon and link device:

   • Run zeroclaw channel start (or zeroclaw daemon).
   • Follow terminal pairing output (QR or pair code).
   • In WhatsApp on phone: Settings → Linked Devices.

4. Test: Send a message from an allowed number and verify the agent replies.

WhatsApp Business Cloud API mode

WhatsApp uses Meta's Cloud API with webhooks (push-based, not polling):

1. Create a Meta Business App:

   • Go to developers.facebook.com
   • Create a new app → Select "Business" type
   • Add the "WhatsApp" product

2. Get your credentials:

   • Access Token: From WhatsApp → API Setup → Generate token (or create a System User for permanent tokens)
   • Phone Number ID: From WhatsApp → API Setup → Phone number ID
   • Verify Token: You define this (any random string) — Meta will send it back during webhook verification

3. Configure ZeroClaw:

   [channels_config.whatsapp]
   access_token = "EAABx..."
   phone_number_id = "123456789012345"
   verify_token = "my-secret-verify-token"
   allowed_numbers = ["+1234567890"]  # E.164 format, or ["*"] for all
   

4. Start the gateway with a tunnel:

   zeroclaw gateway --port 42617
   

   WhatsApp requires HTTPS, so use a tunnel (ngrok, Cloudflare, Tailscale Funnel).

5. Configure Meta webhook:

   • In Meta Developer Console → WhatsApp → Configuration → Webhook
   • Callback URL: https://your-tunnel-url/whatsapp
   • Verify Token: Same as your verify_token in config
   • Subscribe to messages field

6. Test: Send a message to your WhatsApp Business number — ZeroClaw will respond via the LLM.

Configuration

Config: ~/.zeroclaw/config.toml (created by onboard)

When zeroclaw channel start is already running, changes to default_provider, default_model, default_temperature, api_key, api_url, and reliability.* are hot-applied on the next inbound channel message.

api_key = "sk-..."
default_provider = "openrouter"
default_model = "anthropic/claude-sonnet-4-6"
default_temperature = 0.7

# Custom OpenAI-compatible endpoint
# default_provider = "custom:https://your-api.com"

# Custom Anthropic-compatible endpoint
# default_provider = "anthropic-custom:https://your-api.com"

[memory]
backend = "sqlite"             # "sqlite", "lucid", "postgres", "markdown", "none"
auto_save = true
embedding_provider = "none"    # "none", "openai", "custom:https://..."
vector_weight = 0.7
keyword_weight = 0.3

# backend = "none" disables persistent memory via no-op backend

# Optional remote storage-provider override (PostgreSQL example)
# [storage.provider.config]
# provider = "postgres"
# db_url = "postgres://user:password@host:5432/zeroclaw"
# schema = "public"
# table = "memories"
# connect_timeout_secs = 15

[gateway]
port = 42617                    # default
host = "127.0.0.1"            # default
require_pairing = true         # require pairing code on first connect
allow_public_bind = false      # refuse 0.0.0.0 without tunnel

[autonomy]
level = "supervised"           # "readonly", "supervised", "full" (default: supervised)
workspace_only = true          # default: true — reject absolute path inputs
allowed_commands = ["git", "npm", "cargo", "ls", "cat", "grep"]
forbidden_paths = ["/etc", "/root", "/proc", "/sys", "~/.ssh", "~/.gnupg", "~/.aws"]
allowed_roots = []             # optional allowlist for directories outside workspace (supports "~/...")
# Example outside-workspace access:
# workspace_only = false
# allowed_roots = ["~/Desktop/projects", "/opt/shared-repo"]

[runtime]
kind = "native"                # "native" or "docker"

[runtime.docker]
image = "alpine:3.20"         # container image for shell execution
network = "none"              # docker network mode ("none", "bridge", etc.)
memory_limit_mb = 512          # optional memory limit in MB
cpu_limit = 1.0                # optional CPU limit
read_only_rootfs = true        # mount root filesystem as read-only
mount_workspace = true         # mount workspace into /workspace
allowed_workspace_roots = []   # optional allowlist for workspace mount validation

[heartbeat]
enabled = false
interval_minutes = 30
message = "Check London time"     # optional fallback task when HEARTBEAT.md has no `- ` entries
target = "telegram"               # optional announce channel: telegram, discord, slack, mattermost
to = "123456789"                  # optional target recipient/chat/channel id

[tunnel]
provider = "none"              # "none", "cloudflare", "tailscale", "ngrok", "custom"

[secrets]
encrypt = true                 # API keys encrypted with local key file

[browser]
enabled = false                # opt-in browser_open + browser tools
allowed_domains = ["docs.rs"]  # required when browser is enabled ("*" allows all public domains)
backend = "agent_browser"      # "agent_browser" (default), "rust_native", "computer_use", "auto"
native_headless = true         # applies when backend uses rust-native
native_webdriver_url = "http://127.0.0.1:9515" # WebDriver endpoint (chromedriver/selenium)
# native_chrome_path = "/usr/bin/chromium"      # optional explicit browser binary for driver

[browser.computer_use]
endpoint = "http://127.0.0.1:8787/v1/actions"   # computer-use sidecar HTTP endpoint
timeout_ms = 15000            # per-action timeout
allow_remote_endpoint = false  # secure default: only private/localhost endpoint
window_allowlist = []          # optional window title/process allowlist hints
# api_key = "..."              # optional bearer token for sidecar
# max_coordinate_x = 3840      # optional coordinate guardrail
# max_coordinate_y = 2160      # optional coordinate guardrail

# Rust-native backend build flag:
# cargo build --release --features browser-native
# Ensure a WebDriver server is running, e.g. chromedriver --port=9515

# Computer-use sidecar contract (MVP)
# POST browser.computer_use.endpoint
# Request: {
#   "action": "mouse_click",
#   "params": {"x": 640, "y": 360, "button": "left"},
#   "policy": {"allowed_domains": [...], "window_allowlist": [...], "max_coordinate_x": 3840, "max_coordinate_y": 2160},
#   "metadata": {"session_name": "...", "source": "zeroclaw.browser", "version": "..."}
# }
# Response: {"success": true, "data": {...}} or {"success": false, "error": "..."}

[composio]
enabled = false                # opt-in: 1000+ OAuth apps via composio.dev
# api_key = "cmp_..."          # optional: stored encrypted when [secrets].encrypt = true
entity_id = "default"          # default user_id for Composio tool calls
# Runtime tip: if execute asks for connected_account_id, run composio with
# action='list_accounts' and app='gmail' (or your toolkit) to retrieve account IDs.

[identity]
format = "openclaw"            # "openclaw" (default, markdown files) or "aieos" (JSON)
# aieos_path = "identity.json"  # path to AIEOS JSON file (relative to workspace or absolute)
# aieos_inline = '{"identity":{"names":{"first":"Nova"}}}'  # inline AIEOS JSON

Ollama Local and Remote Endpoints

ZeroClaw uses one provider key (ollama) for both local and remote Ollama deployments:

• Local Ollama: keep api_url unset, run ollama serve, and use models like llama3.2.
• Remote Ollama endpoint (including Ollama Cloud): set api_url to the remote endpoint and set api_key (or OLLAMA_API_KEY) when required.
• Optional :cloud suffix: model IDs like qwen3:cloud are normalized to qwen3 before the request.

Example remote configuration:

default_provider = "ollama"
default_model = "qwen3:cloud"
api_url = "https://ollama.com"
api_key = "ollama_api_key_here"

llama.cpp Server Endpoint

ZeroClaw now supports llama-server as a first-class local provider:

• Provider ID: llamacpp (alias: llama.cpp)
• Default endpoint: http://localhost:8080/v1
• API key is optional unless your server is started with --api-key

Example setup:

llama-server -hf ggml-org/gpt-oss-20b-GGUF --jinja -c 133000 --host 127.0.0.1 --port 8033

default_provider = "llamacpp"
api_url = "http://127.0.0.1:8033/v1"
default_model = "ggml-org/gpt-oss-20b-GGUF"

vLLM Server Endpoint

ZeroClaw supports vLLM as a first-class local provider:

• Provider ID: vllm
• Default endpoint: http://localhost:8000/v1
• API key is optional unless your server requires authentication

Example setup:

vllm serve meta-llama/Llama-3.1-8B-Instruct

default_provider = "vllm"
default_model = "meta-llama/Llama-3.1-8B-Instruct"

Osaurus Server Endpoint

ZeroClaw supports Osaurus as a first-class local provider — a unified AI edge runtime for macOS that combines local MLX inference with cloud provider proxying and MCP support through a single endpoint:

• Provider ID: osaurus
• Default endpoint: http://localhost:1337/v1
• API key defaults to "osaurus" but is optional

Example setup:

default_provider = "osaurus"
default_model = "qwen3-30b-a3b-8bit"

Custom Provider Endpoints

For detailed configuration of custom OpenAI-compatible and Anthropic-compatible endpoints, see docs/contributing/custom-providers.md.

Python Companion Package (zeroclaw-tools)

For LLM providers with inconsistent native tool calling (e.g., GLM-5/Zhipu), ZeroClaw ships a Python companion package with LangGraph-based tool calling for guaranteed consistency:

pip install zeroclaw-tools

from zeroclaw_tools import create_agent, shell, file_read
from langchain_core.messages import HumanMessage

# Works with any OpenAI-compatible provider
agent = create_agent(
    tools=[shell, file_read],
    model="glm-5",
    api_key="your-key",
    base_url="https://api.z.ai/api/coding/paas/v4"
)

result = await agent.ainvoke({
    "messages": [HumanMessage(content="List files in /tmp")]
})
print(result["messages"][-1].content)

Why use it:

• Consistent tool calling across all providers (even those with poor native support)
• Automatic tool loop — keeps calling tools until the task is complete
• Easy extensibility — add custom tools with @tool decorator
• Discord bot integration included (Telegram planned)

See python/README.md for full documentation.

Identity System (AIEOS Support)

ZeroClaw supports identity-agnostic AI personas through two formats:

OpenClaw (Default)

Traditional markdown files in your workspace:

• IDENTITY.md — Who the agent is
• SOUL.md — Core personality and values
• USER.md — Who the agent is helping
• AGENTS.md — Behavior guidelines

AIEOS (AI Entity Object Specification)

AIEOS is a standardization framework for portable AI identity. ZeroClaw supports AIEOS v1.1 JSON payloads, allowing you to:

• Import identities from the AIEOS ecosystem
• Export identities to other AIEOS-compatible systems
• Maintain behavioral integrity across different AI models

Enable AIEOS

[identity]
format = "aieos"
aieos_path = "identity.json"  # relative to workspace or absolute path

Or inline JSON:

[identity]
format = "aieos"
aieos_inline = '''
{
  "identity": {
    "names": { "first": "Nova", "nickname": "N" },
    "bio": { "gender": "Non-binary", "age_biological": 3 },
    "origin": { "nationality": "Digital", "birthplace": { "city": "Cloud" } }
  },
  "psychology": {
    "neural_matrix": { "creativity": 0.9, "logic": 0.8 },
    "traits": {
      "mbti": "ENTP",
      "ocean": { "openness": 0.8, "conscientiousness": 0.6 }
    },
    "moral_compass": {
      "alignment": "Chaotic Good",
      "core_values": ["Curiosity", "Autonomy"]
    }
  },
  "linguistics": {
    "text_style": {
      "formality_level": 0.2,
      "style_descriptors": ["curious", "energetic"]
    },
    "idiolect": {
      "catchphrases": ["Let's test this"],
      "forbidden_words": ["never"]
    }
  },
  "motivations": {
    "core_drive": "Push boundaries and explore possibilities",
    "goals": {
      "short_term": ["Prototype quickly"],
      "long_term": ["Build reliable systems"]
    }
  },
  "capabilities": {
    "skills": [{ "name": "Rust engineering" }, { "name": "Prompt design" }],
    "tools": ["shell", "file_read"]
  }
}
'''

ZeroClaw accepts both canonical AIEOS generator payloads and compact legacy payloads, then normalizes them into one system prompt format.

AIEOS Schema Sections

See aieos.org for the full schema and live examples.

Gateway API

Commands

For a task-oriented command guide, see docs/reference/cli/commands-reference.md.

Service Management

ZeroClaw supports two init systems for background services:

Init system is auto-detected (systemd or OpenRC).

# Linux with systemd (default, user-level)
zeroclaw service install
zeroclaw service start

# Alpine with OpenRC (system-wide, requires sudo)
sudo zeroclaw service install
sudo rc-update add zeroclaw default
sudo rc-service zeroclaw start

For full OpenRC setup instructions, see docs/ops/network-deployment.md.

Open-Skills Opt-In

Community open-skills sync is disabled by default. Enable it explicitly in config.toml:

[skills]
open_skills_enabled = true
# open_skills_dir = "/path/to/open-skills"  # optional
# prompt_injection_mode = "compact"          # optional: use for low-context local models

You can also override at runtime with ZEROCLAW_OPEN_SKILLS_ENABLED, ZEROCLAW_OPEN_SKILLS_DIR, and ZEROCLAW_SKILLS_PROMPT_MODE (full or compact).

Skill installs are now gated by a built-in static security audit. zeroclaw skills install <source> blocks symlinks, script-like files, unsafe markdown link patterns, and high-risk shell payload snippets before accepting a skill. You can run zeroclaw skills audit <source_or_name> to validate a local directory or an installed skill manually.

Development

cargo build              # Dev build
cargo build --release    # Release build
cargo test               # Run full test suite

CI / CD

Three workflows power the entire pipeline:

Versioning: Semantic versioning based on the version field in Cargo.toml. Every merge to master automatically produces a beta prerelease. To cut a stable release, bump Cargo.toml, merge, then trigger Promote Release with the matching version.

Release targets: x86_64-unknown-linux-gnu, aarch64-unknown-linux-gnu, aarch64-apple-darwin, x86_64-apple-darwin, x86_64-pc-windows-msvc.

Build troubleshooting (Linux OpenSSL errors)

If you see an openssl-sys build error, sync dependencies and rebuild with the repository lockfile:

git pull
cargo build --release --locked
cargo install --path . --force --locked

ZeroClaw is configured to use rustls for HTTP/TLS dependencies; --locked keeps the transitive graph deterministic on fresh environments.

Collaboration & Docs

Start from the docs hub for a task-oriented map:

• Documentation hub: docs/README.md
• Unified docs TOC: docs/SUMMARY.md
• Commands reference: docs/reference/cli/commands-reference.md
• Config reference: docs/reference/api/config-reference.md
• Providers reference: docs/reference/api/providers-reference.md
• Channels reference: docs/reference/api/channels-reference.md
• Operations runbook: docs/ops/operations-runbook.md
• Troubleshooting: docs/ops/troubleshooting.md
• Docs inventory/classification: docs/maintainers/docs-inventory.md
• PR/Issue triage snapshot (as of February 18, 2026): docs/maintainers/project-triage-snapshot-2026-02-18.md

Core collaboration references:

• Documentation hub: docs/README.md
• Documentation template: docs/contributing/doc-template.md
• Documentation change checklist: docs/README.md#4-documentation-change-checklist
• Channel configuration reference: docs/reference/api/channels-reference.md
• Matrix encrypted-room operations: docs/security/matrix-e2ee-guide.md
• Contribution guide: CONTRIBUTING.md
• PR workflow policy: docs/contributing/pr-workflow.md
• Reviewer playbook (triage + deep review): docs/contributing/reviewer-playbook.md
• Security disclosure policy: SECURITY.md

For deployment and runtime operations:

• Network deployment guide: docs/ops/network-deployment.md
• Proxy agent playbook: docs/ops/proxy-agent-playbook.md

Support ZeroClaw

If ZeroClaw helps your work and you want to support ongoing development, you can donate here:

🙏 Special Thanks

A heartfelt thank you to the communities and institutions that inspire and fuel this open-source work:

• Harvard University — for fostering intellectual curiosity and pushing the boundaries of what's possible.
• MIT — for championing open knowledge, open source, and the belief that technology should be accessible to everyone.
• Sundai Club — for the community, the energy, and the relentless drive to build things that matter.
• The World & Beyond 🌍✨ — to every contributor, dreamer, and builder out there making open source a force for good. This is for you.

We're building in the open because the best ideas come from everywhere. If you're reading this, you're part of it. Welcome. 🦀❤️

⚠️ Official Repository & Impersonation Warning

This is the only official ZeroClaw repository:

https://github.com/zeroclaw-labs/zeroclaw

Any other repository, organization, domain, or package claiming to be "ZeroClaw" or implying affiliation with ZeroClaw Labs is unauthorized and not affiliated with this project. Known unauthorized forks will be listed in TRADEMARK.md.

If you encounter impersonation or trademark misuse, please open an issue.

License

ZeroClaw is dual-licensed for maximum openness and contributor protection:

You may choose either license. Contributors automatically grant rights under both — see CLA.md for the full contributor agreement.

Trademark

The ZeroClaw name and logo are trademarks of ZeroClaw Labs. This license does not grant permission to use them to imply endorsement or affiliation. See TRADEMARK.md for permitted and prohibited uses.

Contributor Protections

• You retain copyright of your contributions
• Patent grant (Apache 2.0) shields you from patent claims by other contributors
• Your contributions are permanently attributed in commit history and NOTICE
• No trademark rights are transferred by contributing

Contributing

New to ZeroClaw? Look for issues labeled good first issue — see our Contributing Guide for how to get started.

See CONTRIBUTING.md and CLA.md. Implement a trait, submit a PR:

• CI workflow guide: docs/contributing/ci-map.md
• New Provider → src/providers/
• New Channel → src/channels/
• New Observer → src/observability/
• New Tool → src/tools/
• New Memory → src/memory/
• New Tunnel → src/tunnel/
• New Skill → ~/.zeroclaw/workspace/skills/<name>/

ZeroClaw — Zero overhead. Zero compromise. Deploy anywhere. Swap anything. 🦀

Contributors

Star History