FastClaw

A lightweight but powerful AI Agent framework in Python.

PyPI Installation (Recommended)

# Install
pip install fastclaw-ai

# Run
fastclaw start

# Persistent running (recommended for server)
nohup fastclaw start > /tmp/fastclaw.log 2>&1 &

Configure Agent

After starting, visit http://localhost:8765 , click Settings in the top right corner to configure main_agent, recommended: deepseek-chat.

Or edit the config file:

vim ~/.fastclaw/workspace/data/agents/main_agent/metadata.json

CLI Usage

# Interactive chat
fastclaw chat

# New session
fastclaw chat --new

# Check status
fastclaw status

Git Clone

git clone https://github.com/kandada/fastclaw.git
cd fastclaw
pip install -r requirements.txt

# Run
python main.py start

# Persistent running
nohup python main.py start > fastclaw.log 2>&1 &

Configure Agent

After starting, visit http://localhost:8765 , click Settings in the top right corner to configure main_agent, recommended: deepseek-chat.

Or edit the config file:

vim workspace/data/agents/main_agent/metadata.json

CLI Usage

# Interactive chat
python main.py chat

# New session
python main.py chat --new

# Check status
python main.py status

Design Philosophy

FastClaw is built on the core concept of event-driven agent orchestration. Key design principles:

• Graph-based Agent: Uses a directed graph to orchestrate agent behavior, supporting conditional branching between nodes (e.g., tool execution, response generation)
• Event Streaming: Real-time streaming output of LLM responses for better user experience
• Tool System: Extensible tool framework supporting Shell commands, skills, and custom integrations
• Session Management: Persistent conversation history with multi-session support
• Cron Scheduling: Built-in cron-style task scheduling for automated workflows

Features

• 🤖 LLM-powered - Built on FastMind framework with streaming support
• 🔧 Tool Calling - Execute Shell commands, skills, and more
• ⏰ Cron Jobs - Schedule tasks with cron expressions
• 💬 Multi-channel - Feishu, iMessage integrations
• 🎨 Extensible - Easy to add custom skills and agents
• 🐍 Python Ecosystem - Seamlessly call professional libraries like numpy and pandas, enabling AI to use Python ecosystem tools as skillfully as humans

Changelog

v1.1.25 (2026-08-30)

• ✨ New: Default context (max_tokens / unload_threshold_tokens) raised to 256000 — most mainstream LLMs now support this window or larger
• 🔧 Improvement: Default agent personality files (SOUL.md / USER.md / AGENT.md) translated to English; USER.md language preference now defaults to English for international users

v1.1.24 (2026-08-29)

• ✨ New: Full Anthropic gateway support — Messages API streaming, tool calling, extended thinking (adaptive → enabled → none state machine), and auto base_url adjustment for MiniMax / DeepSeek / Moonshot Anthropic-compatible endpoints
• ✨ New: OpenAI gateway thinking compatibility — unified extraction of reasoning_content / thinking field / <think>...</think> tags (MiniMax-M3, Qwen3, etc.)
• 🐛 Fix: WebUI in-progress view no longer drops tool calls — repeated same-name tools across ReAct rounds and parallel calls now each get their own card (id-based matching)
• 🐛 Fix: WebUI message block ordering — content now appears before tool_call in history replay and resume, consistent with live streaming (thinking → content → tool_call → tool_result)
• 🐛 Fix: WebUI tool result cards unified to a generic "tool" label, consistent with the stored message format

v1.1.23 (2026-08-15)

• ✨ New: Cron jobs now support Feishu (Lark) — scheduled tasks can deliver results to Feishu chats
• 🔧 Improvement: WebUI refactor — optimized conversation list interaction
• 🐛 Fix: Cron scheduler fixes and regression fixes

v1.1.22 (2026-07-22)

• ✨ New: skill_creator bundled meta-skill — guides the agent to create, update and optimize skills via chat
• ✨ New: Skills management tab in WebUI — create and edit user skills with Remote Endpoint / Secret support
• ✨ New: Skills CRUD API endpoints (/api/skills/user) — create, read, update, delete user skills
• ✨ New: Hot-reload for skills — new skills are auto-registered at runtime, no restart needed
• ✨ New: "Open Skills Dir" button in WebUI to open the skills directory in file explorer
• 🔧 Improvement: All bundled skill documentation (SKILL.md) translated to English
• 🔧 Improvement: Template _template excluded from skill lists
• 🐛 Fix: Skills list in system prompt now updates when new skills are created

v1.1.21 (2026-07-09)

• 🐛 Fix: Added python-socks[asyncio] dependency to fix Lark (Feishu) connection failure when using a SOCKS proxy

v1.1.20 (2026-07-03)

v1.1.19 (2026-06-29)

• 🐛 Fix: Resolved SyntaxError on Python 3.10/3.11 caused by backslash (\) inside f-string {} expressions (a Python 3.12+ syntax)
• 🔧 Improvement: Added version compatibility comments at f-string usage sites to prevent future regressions

v1.1.18 (2026-06-27)

• 🔧 Improvement: Added copyright and GPLv3 license headers to all source files

v1.1.17 (2026-06-27)

• 🐛 Fix: Resolved UnicodeDecodeError on Windows Chinese systems by adding encoding="utf-8" to all read_text() / write_text() calls, fixing the crash on second startup
• 🐛 Fix: WebUI agent edit/create buttons now show error messages on failure instead of silently doing nothing
• 🐛 Fix: Removed 3 extra </div> tags from WebUI HTML that could cause DOM parsing issues
• 🔧 Improvement: Startup messages now show http://localhost:{port} alongside http://0.0.0.0:{port} for easier access on Windows
• 🔧 Improvement: Added diagnostic logging to silent except blocks in router.py for easier debugging

v1.1.16 (2026-06-26)

• 🔧 Improvement: Made ripgrep an optional dependency to avoid compilation issues on Windows without MSVC
• 🔧 Improvement: Added Python 3.13 / 3.14 support

v1.1.15 (2026-06-12)

v1.1.14 (2026-06-12)

v1.1.13 (2026-06-12)

v1.1.12 (2026-06-11)

• 🐛 Fix: WebUI bugs and cron scheduler bugs

v1.1.11 (2026-06-02)

v1.1.10 (2026-05-31)

v1.1.9 (2026-05-31)

v1.1.8 (2026-05-25)

• ✨ Feature: LLM Gateway selector (OpenAI / Anthropic) in WebUI agent settings
• ✨ Feature: Auto-bootstrap default agents from GitHub when workspace is empty
• ✨ Feature: Auto-generate settings.json with defaults when missing
• 🔧 Improvement: Increased default Unload Threshold from 80k to 156k

v1.1.7 (2026-05-21)

• 🐛 Fix: Context unloading now only affects LLM input, messages.jsonl always keeps full history
• 🐛 Fix: Message timestamps are no longer overwritten on every save
• 🐛 Fix: CancelledError no longer triggers duplicate response generation
• 🧪 Test: Added context unloading, sliding window, and timestamp preservation tests

License

GPL-3.0

Copyright (c) 2024-2026 xiefujin. All rights reserved.

Official Website

https://www.fastclaw.world/ | https://fastclaw-ai.com/

FastMind Ecosystem

FastClaw is part of the FastMind ecosystem. Other projects built on FastMind:

• FastClaw — A lightweight, general-purpose AI Agent framework with event-driven agent orchestration, graph-based agent design, and extensible tool system
• FastBot — A dual-loop embodied intelligence robot simulation demo, exploring perception-action closed-loop control with LLM-driven decision making

Acknowledgments

Inspired by OpenClaw. Special thanks to the open source community.

FastClaw is powerful, but its use depends entirely on the user - all responsibility lies with the user yourself.

Author：xiefujin email: 490021684@qq.com，welcome to contact me.