Claude, GPT, Gemini, DeepSeek, Ollama, or your own endpoint. One tool, every model, fully hackable in Python.
A complete AI coding assistant that runs in your terminal, works with any model, and extends with plugins.
Claude, GPT-4o, Gemini, DeepSeek, Qwen, Kimi, MiniMax, Ollama, or any OpenAI-compatible endpoint. Switch with /model.
Read, Write, Edit, Bash, Glob, Grep, WebFetch, WebBrowse, ReadPDF, ReadEmail, NotebookEdit, and more — all callable by the AI.
Spawn specialized sub-agents (coder, reviewer, researcher) in parallel with git worktree isolation. Background execution with named addressing.
Control from your phone via Telegram, WeChat, or Slack bridges. Job queues, slash commands, and live tool tracking.
Permission system (auto/manual/plan modes), prompt injection detection, credential filtering in /config, safe stdio wrapper.
Launch research assistants, bug fixers, paper writers, and auto-coders that run independently. 4 built-in templates + custom.
Record, transcribe, auto-submit. Offline Whisper — no API key, no subscription. 99 languages.
Read PDFs, OCR images, parse spreadsheets, read and send emails — all via natural language commands.
Install plugins from git URLs. Extend with custom tools, commands, skills, and MCP servers. Full authoring guide included.
10+ providers, 200+ models. Cloud or local — your choice.
Real workflows, real productivity gains.
# Free, private, nothing leaves your machine $ cheetahclaws --model ollama/qwen2.5-coder [project] 12% » Review src/api.py for security issues and performance problems. I'll analyze the file... [Read] src/api.py (142 lines) [Grep] SQL injection patterns... Found 3 issues: 1. Line 42: SQL injection via f-string 2. Line 78: Missing rate limiting 3. Line 95: Unvalidated redirect
Run entirely offline with Ollama. Your code never leaves your machine. The AI reads files, searches for patterns, and gives actionable feedback.
Control CheetahClaws from your phone while it runs on your workstation. Send commands, get results, manage tasks.
# You (Telegram): What files changed in the last commit? # CheetahClaws: [Bash] git log --oneline -1 [Bash] git diff HEAD~1 --stat Last commit: "fix auth middleware" 3 files changed: - src/auth.py (+12 -5) - tests/test_auth.py (+28) - config.yaml (+1) # You: Fix the type error on line 42 # CheetahClaws edits the file...
[project] » Read the contract at ~/Documents/contract.pdf and summarize the key payment terms. [ReadPDF] contract.pdf (24 pages) Key payment terms: 1. Net 30 payment cycle 2. 2% early payment discount 3. Late fee: 1.5% monthly [project] » Check my latest emails from [email protected] [ReadEmail] 3 emails found 1. "Invoice #4521" - Apr 14 2. "Q1 Report attached" - Apr 10 3. "Budget approval" - Apr 8
Read PDFs, parse Excel spreadsheets, OCR images, read and send emails — all through natural language.
See the full range of capabilities.
Full browser chat — sessions sidebar, tool cards, approval prompts, markdown streaming
Spawn expert personas to debate and synthesize ideas
Agent works proactively in the background without prompts
14 workflow shortcuts: brainstorm, review, commit, and more
Control your agent from your phone — send tasks, get results
Control your agent from WeChat (微信) — scan QR, start chatting
Control your agent from a Slack channel — team-friendly
Analyze codebase in read-only mode, write a plan, then implement
Spawn parallel sub-agents with git worktree isolation
Save and recall facts, preferences, and decisions across sessions
Switch between Claude, GPT, Gemini, Ollama mid-conversation
Auto-snapshot every turn — rewind files and conversation to any point
Speak your prompt — offline Whisper transcription, 99 languages
Paste an image and ask questions — works with local and cloud models
Generate videos: story → TTS narration → AI images → subtitles → MP4
Automatically implement TODO items from brainstorm output
Split panes, run long tasks, capture output — 11 tmux tools
Backup conversations to GitHub Gist — resume from any machine
Autonomous trading agent — monitor markets, analyze signals, execute strategies
Run any shell command inline with !command — git, ls, python, etc.
All three are personal AI assistants — but they aim at different goals. Here's where CheetahClaws stands out vs. OpenClaw and Claude Code.
| CheetahClaws | OpenClaw | Claude Code | |
|---|---|---|---|
| Primary focus | Coding & productivity assistant for terminal + web | Multi-channel messaging assistant (WhatsApp, iMessage, Slack…) | Coding assistant tightly coupled to Claude API |
| Language & runtime | Python 3.10–3.13, ~40K LOC, fully readable | TypeScript / Node.js 22+ | TypeScript / Node.js (compiled, ~12 MB minified) |
| Model providers | ✓ 10+ providers, 200+ models — Claude, GPT, Gemini, DeepSeek, Qwen, Kimi, GLM, MiniMax, Ollama, vLLM, LM Studio | Cloud-first, primarily one frontier provider | ✗ Anthropic Claude only |
| Local / offline models | ✓ Native Ollama / vLLM / LM Studio with streaming & tool-calling | Limited; cloud-first design | ✗ None — requires Anthropic API |
| Switch model mid-chat | ✓ /model swaps any provider live |
Limited | ✗ Claude variants only |
| Built-in tools | ✓ 33 tools — Read, Write, Edit, Bash, Glob, Grep, WebFetch, WebBrowse, ReadPDF, ReadEmail, NotebookEdit… | Channel & messaging tools | ~15 core file/shell/web tools |
| Remote control (phone/chat) | ✓ Telegram, WeChat, Slack bridges with job queue | ✓ 22+ channels (its core strength) | ✗ Terminal / IDE only |
| Voice input (offline) | ✓ Offline Whisper, 99 languages, no API key | Speak / listen on macOS / iOS / Android | ✗ Not built-in |
| PDF / Excel / Email tools | ✓ ReadPDF, OCR, Excel, IMAP / SMTP — all native | Via channel integrations | ✗ Bring your own |
| Multi-agent & sub-agents | ✓ Parallel sub-agents with git-worktree isolation, 4 autonomous templates | Limited | ✓ Task / sub-agent system |
| Hackability | ✓ Plain Python — read it, fork it, write a plugin in an afternoon | TypeScript monorepo, larger surface area | ✗ Closed source binary; source is reverse-engineered |
| License | Apache 2.0 — fully open | MIT — fully open | Proprietary (Anthropic) |
| Cost | Free — pay only your model provider (or $0 with Ollama) | Free + your model API | Bundled with Anthropic API / subscription |
Pick any model, run anywhere, hack the source. The fastest path from "I want an AI to do X" to actually doing it.
A messaging-first assistant. Great if you want one bot answering you across 22+ chat platforms.
Anthropic's official terminal coding agent. Polished, but locked to one vendor and one language.
One command to install. First run guides you through setup. No build step, no daemon, no subscription.