An agentic AI assistant for chat surfaces, MicroClaw offers a channel-agnostic core with adapters for Telegram, Discord, Slack, Feishu/Lark, IRC, and Web. It empowers users with advanced AI capabilities, including full tool execution, persistent memory, and multi-LLM provider support, streamlining complex tasks and integrating AI into conversational workflows.

How It Works

Each message initiates an agentic loop where the LLM can execute tools, analyze results, and perform multi-step reasoning, with a default limit of 100 iterations per request. MicroClaw employs a channel-agnostic architecture, leveraging platform adapters for diverse chat services and supporting multiple LLM providers like Anthropic and OpenAI-compatible APIs. Key features include session resume, context compaction for managing long conversations, and a robust tool system for interacting with the environment.

Quick Start & Requirements

Installation is streamlined via a one-line installer script (curl -fsSL https://microclaw.ai/install.sh | bash) or Homebrew (brew install microclaw). Building from source requires Rust. First-time setup is guided by an interactive wizard (microclaw setup), which prompts for necessary channel bot credentials (e.g., Telegram, Discord) and LLM API keys. Docker is required for the optional sandbox mode. Official documentation and installers are available at https://microclaw.ai/.

Highlighted Details

• Multi-Platform Integration: Supports Telegram, Discord, Slack, Feishu/Lark, IRC, and Web interfaces.
• Extensive Toolset: Includes capabilities for executing shell commands, file I/O, web searching and fetching, scheduling tasks, and delegating sub-tasks to parallel agents.
• Persistent Memory: Manages global and per-chat memory via AGENTS.md files and structured SQLite storage, with optional semantic KNN retrieval.
• Advanced Context Management: Features session resume, context compaction for long conversations, and a continuous typing indicator.
• Extensible Skills: Supports a modular skill system compatible with Anthropic's standard for specialized task execution.
• Local Web UI: Provides a cross-channel history interface for reviewing and managing conversations.
• Docker Sandbox: Offers an optional Docker sandbox for secure execution of shell commands.

Maintenance & Community

The project is noted as being under active development, with an invitation for contributions. Specific community channels or detailed maintenance logs are not explicitly detailed in the provided README.

Licensing & Compatibility

MicroClaw is released under the MIT license, which is permissive and generally suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

As the project is under active development, features are subject to change. Certain advanced functionalities, such as semantic memory retrieval, require building the project with specific features enabled (--features sqlite-vec). The Docker sandbox mode necessitates a functional Docker installation.