<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> GoClaw is a multi-agent AI gateway designed to connect Large Language Models (LLMs) with tools, channels, and data. It addresses the need for robust, production-grade orchestration of AI agents, offering advanced features like team collaboration, delegation, and quality-gated workflows. The project targets developers and power users seeking a scalable, secure, and easily deployable AI infrastructure solution.

How It Works

GoClaw is engineered as a single, static Go binary, minimizing runtime dependencies and enabling rapid startup (<1s) even on low-cost hardware ($5 VPS). Its core innovation lies in its production-grade security features and a PostgreSQL backend supporting true database-native multi-tenancy for per-user workspaces and isolated sessions. It orchestrates complex multi-agent interactions through patterns like agent teams with shared task boards, inter-agent delegation (sync/async), conversation handoff, and evaluate-loop quality gates.

Quick Start & Requirements

Installation can be done by building from source (make build, ./goclaw onboard) or via Docker Compose. Prerequisites include Go 1.25+ and PostgreSQL 18 with pgvector for managed mode. The project offers an interactive setup wizard (./goclaw onboard) and an auto-onboarding feature via environment variables for seamless integration in CI/CD pipelines. Docker Compose files are provided for various deployment scenarios, including standalone (file-based) and managed (PostgreSQL) modes, with optional overlays for OpenTelemetry tracing and Tailscale remote access.

Highlighted Details

• LLM Providers: Supports 13+ providers including Anthropic (native HTTP+SSE with prompt caching), OpenAI, OpenRouter, Groq, Gemini, Mistral, and more.
• Messaging Channels: Integrates with Telegram, Discord, Zalo, Feishu/Lark, and WhatsApp.
• Security: Features a 5-layer defense including rate limiting, prompt injection detection, SSRF protection, shell deny patterns, and AES-256-GCM encryption for API keys in managed mode.
• Agent Orchestration: Advanced patterns include agent teams with task boards and mailboxes, sync/async delegation with permission links, conversation handoff, generator-evaluator feedback loops (evaluate loop), and hook-based quality gates.
• Deployment: Single binary (~25MB base) with minimal resource footprint (~35 MB RAM idle).
• Multi-tenancy: DB-native multi-tenancy via PostgreSQL for per-user workspaces and isolated sessions.

Maintenance & Community

The project acknowledges its foundation in the original OpenClaw project. No specific details regarding active maintainers, community channels (like Discord or Slack), or sponsorship were found in the provided README.

Licensing & Compatibility

GoClaw is released under the MIT license. This permissive license generally allows for commercial use, modification, and distribution, including integration within closed-source applications.

Limitations & Caveats

While many advanced features like agent delegation, teams, handoff, evaluate loops, quality gates, and various messaging channel integrations are implemented, they are noted as "Implemented but Not Fully Tested" and require end-to-end validation in production workflows. The standalone mode lacks per-user isolation and advanced orchestration capabilities, making the managed PostgreSQL mode the recommended setup for unlocking the full feature set.