zeroclaw-labs/zeroclaw

⭐ 31,208  ·  Rust  ·  GitHub Repo

Fast, small, and fully autonomous AI personal assistant infrastructure, ANY OS, ANY PLATFORM — deploy anywhere, swap anything 🦀

agent agentic ai infra ml openclaw os zeroclaw

1-Sentence Summary

Your own fully autonomous AI agent: self-hosted, single binary, any OS, any platform, total data ownership.

🔥 Key Capabilities & USP

• Multi-Channel Communication: One agent loop simultaneously serves 30+ channels (Discord, Telegram, Matrix, email, CLI, webhooks). Pain point solved: No more managing separate bots per platform or missing messages across fragmented communication tools.
• Provider-Agnostic Model Support: Pluggable LLM providers (Anthropic, OpenAI, Ollama, ~20 others) with automatic fallback chains and routing. Pain point solved: Eliminates vendor lock-in and single-point-of-failure; your agent keeps running even if one provider goes down.
• Security-First Design: Default supervised autonomy with approval gates, workspace boundaries, OS-level sandboxes (Landlock/Bubblewrap/Seatbelt/Docker), and cryptographic receipts for every action. Pain point solved: Trust your agent to execute actions without fear of unintended system modifications or data leaks.
• Hardware Capability: Direct GPIO/I2C/SPI/USB control on Raspberry Pi, STM32, Arduino, and ESP32 via the Peripheral trait. Pain point solved: Bridges the gap between software agents and physical world automation for robotics and IoT projects.
• SOP Engine & Gateway: Event-triggered Standard Operating Procedures (MQTT/webhook/cron/peripheral) with approval gates, plus an HTTP/WebSocket gateway with a web dashboard. Pain point solved: Automate complex multi-step workflows without writing glue code, while maintaining human oversight.

USP: ZeroClaw is the only fully autonomous, self-hosted AI agent infrastructure that runs as a single Rust binary on any OS or platform, giving you complete ownership of your agent, data, and compute.

Technical Architecture

Layer	Components	Description
Channels	30+ adapters (Discord, Telegram, Matrix, email, webhooks, CLI)	Ingest messages from any source into the same agent loop
Gateway	REST API + WebSocket + Web Dashboard	Chat, memory, config, and tool management UI
ACP	JSON-RPC 2.0 protocol	Standardized agent communication protocol for external integrations
Runtime	Agent loop + Security policy + SOP engine	Core orchestration with approval gates and event triggers
Providers	Anthropic, OpenAI, Ollama, ~20 others	Pluggable LLM backends with fallback chains
Tools	Filesystem, shell, GPIO, web, etc.	Extensible tool system with cryptographic receipts
Memory	SQLite + Embeddings	Persistent state and vector search for context retention

Tech Stack: Rust (edition 2024), single binary deployment, modular plugin architecture, OS-level sandboxing.

Quick Start Guide

# Install with one command
curl -fsSL https://raw.githubusercontent.com/zeroclaw-labs/zeroclaw/master/install.sh | bash

# Or clone and run
git clone https://github.com/zeroclaw-labs/zeroclaw.git
cd zeroclaw
./install.sh

# Installation options
./install.sh --prebuilt              # always prebuilt; don't ask
./install.sh --source                # always build from source
./install.sh --minimal               # kernel only (~6.6 MB)
./install.sh --source --features agent-runtime,channel-discord  # custom feature set
./install.sh --skip-onboard          # install only, run `zeroclaw onboard` later
./install.sh --list-features         # print available feature flags

# Quick start
zeroclaw onboard                  # wizard: picks a provider, wires channels
zeroclaw agent                    # interactive chat in the terminal
zeroclaw service install          # register as systemd/launchctl/Windows Service
zeroclaw service start            # run it always-on in the background

Pros, Cons & Use Cases

Pros

• True autonomy: Runs fully self-hosted with no cloud dependency
• Cross-platform: Single binary works on any OS (Linux, macOS, Windows) and any hardware (x86, ARM, RISC-V)
• Security by default: Sandboxed execution, approval gates, and cryptographic receipts
• Extreme flexibility: Swap providers, channels, tools, and memory backends without code changes
• Hardware integration: Direct peripheral control for robotics and IoT

Cons

• Manual configuration: Requires editing a TOML file; no GUI-based setup wizard
• Security friction: Default supervised mode requires approval for every action, which may feel restrictive for casual users
• No managed cloud option: You must provide your own compute and infrastructure
• Learning curve: Understanding the modular architecture and configuration options takes time

Who should NOT use this?

• Casual users who want a plug-and-play AI assistant with zero configuration
• Teams needing a managed SaaS solution with SLAs and support contracts
• Users uncomfortable with CLI tools and configuration files
• Projects requiring cloud-native scaling (auto-scaling, load balancing across instances)

Ideal Use Cases

• Developers building custom AI agents with full control over models, tools, and data
• Privacy-conscious users who want to keep all data on-premise
• IoT/robotics enthusiasts integrating AI with physical hardware (Raspberry Pi, Arduino, ESP32)
• DevOps teams automating infrastructure tasks with secure, auditable agent actions
• Multi-platform power users who need one agent across Discord, Telegram, email, and CLI

Community & Activity

ZeroClaw is on fire with 31,208 stars and a last update on May 10, 2026—this is a rapidly evolving, actively maintained project. The community momentum is extraordinary, indicating strong developer adoption and trust. With a single Rust binary that solves the fragmentation problem of AI assistants, ZeroClaw is positioned to become the de facto standard for self-hosted agent infrastructure. The active development cadence and large star count signal a project that is both stable for production use and continuously improving.