Repolode
Submit
SAIL-Research-Lab

SAIL-Research-Lab/cheetahclaws

A Python-native agent harness and terminal coding assistant that works with many cloud and local LLMs, with tools, memory, skills and a Web UI.

Open-source alternative to:Claude Code

View on GitHubWebsite
Stars
789
Forks
274
Open issues
27
Last commit
10 days ago

What is it?

CheetahClaws is a Python agent harness that runs a tool-using AI assistant from the terminal. It reimplements the core agent loop of Claude Code in a readable codebase, and its model layer lets you switch between Anthropic, OpenAI, Gemini, DeepSeek, Qwen, Kimi, MiniMax and other providers, plus local models served by Ollama, LM Studio, vLLM or any OpenAI-compatible endpoint, using the same --model flag.

The assistant ships with about 28 built-in tools for reading, writing and editing files, running shell commands, searching code, fetching web pages and managing tasks, and it can be extended through MCP servers, git-based plugins and Markdown skills. Persistent memory, context compression, permission prompts with an autonomous mode, checkpoints and session autosave are included. Optional modules add a Web UI with a browser terminal, Telegram, WeChat, Slack and QQ bridges, voice input, and trading and research workflows.

Key features

  • 01Switch between cloud providers and local models with one --model flag.
  • 02Includes about 28 built-in tools for files, shell, search and web access.
  • 03Extensible through MCP servers, git plugins and Markdown skills.
  • 04Persistent user and project memory with context compression.
  • 05Permission prompts with a selectable autonomous mode and a command denylist.
  • 06Optional Web UI with a browser terminal and chat bridges to Telegram, Slack and others.

Use cases

  • Running a terminal coding assistant against a local Ollama or vLLM model without sending code to a cloud API.
  • Comparing how different LLM providers handle the same coding task with one tool.
  • Studying or customizing an agent loop in a small Python codebase.
  • Driving an assistant remotely from a chat app through its bridge modules.

Topics

#agent#llm#ai-agent#claude-code#mcp#multi-model#python#cli#agentic-ai#claude

FAQ

What is cheetahclaws?
A Python-native agent harness and terminal coding assistant that works with many cloud and local LLMs, with tools, memory, skills and a Web UI. CheetahClaws is a Python agent harness that runs a tool-using AI assistant from the terminal. It reimplements the core agent loop of Claude Code in a readable codebase, and its model layer lets you switch between Anthropic, OpenAI, Gemini, DeepSeek, Qwen, Kimi, MiniMax and other providers, plus local models served by Ollama, LM Studio, vLLM or any OpenAI-compatible endpoint, using the same --model flag.
Is cheetahclaws free to use? What license does it have?
cheetahclaws is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is cheetahclaws still maintained?
Actively maintained. The last commit to cheetahclaws was on Oct 12, 2026.
What can I use cheetahclaws for?
Running a terminal coding assistant against a local Ollama or vLLM model without sending code to a cloud API. Comparing how different LLM providers handle the same coding task with one tool. Studying or customizing an agent loop in a small Python codebase. Driving an assistant remotely from a chat app through its bridge modules.
Where can I get cheetahclaws?
The source code and installation instructions are on GitHub: https://github.com/SAIL-Research-Lab/cheetahclaws

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects