Open Source · v1.0.0 · MIT License

OpenClaw AI — Autonomous Agent Framework for Developers

OpenClaw AI lets you build, run, and chain autonomous AI agents locally or in the cloud. Integrate with OpenAI, Claude, Ollama and more — automate any task with natural language commands.

Get Started Installation Guide ⭐ Star on GitHub

⭐ 2.4k stars · 📦 18k+ installs · 🐍 Python 3.10+ · 📄 MIT License

🤖

Autonomous Agents

Define tasks in plain English. OpenClaw decomposes goals into sub-tasks and executes them autonomously, step by step.

🏠

100% Local AI

Run agents fully offline using Ollama or any GGUF model. Your data never leaves your machine. No API key required.

⚡

Multi-LLM Support

Plug in OpenAI GPT-4, Anthropic Claude, Mistral, LLaMA 3, Gemini — switch providers with a single config change.

🔗

Tool Ecosystem

Built-in tools: file system, web search, code execution, browser control, API calls, database queries and more.

🔌

Easy Integrations

Connect to Telegram, Discord, Slack bots. Integrate with CI/CD pipelines, GitHub Actions, and REST APIs.

📦

Open Source

MIT licensed, community-driven. Fork it, extend it, build your own tools. No vendor lock-in ever.

System Requirements 🐍 Python 3.10+ · 💾 8 GB RAM (16 GB recommended) · 📦 pip or conda · 🖥️ Linux · macOS · Windows

Quick Start in 3 Steps

From zero to running your first AI agent in under 5 minutes.

Install OpenClaw

Install via pip. Python 3.10+ required.

pip install openclaw

Set your API key

Configure your LLM provider (or use local Ollama).

openclaw config set api_key YOUR_KEY

Run your first agent

Give your agent a task in plain English.

openclaw run "Summarize all .txt files in ~/docs"

Need more details? Full installation guide → · Getting started tutorial →

✅ Expected output after step 3

🧠 OpenClaw AI v1.0.0
📋 Task: Summarize all .txt files in ~/docs
🔍 Scanning ~/docs... found 4 .txt files

Step 1/3: Reading file: notes.txt (1.2 KB)
Step 2/3: Reading file: todo.txt (0.4 KB)
Step 3/3: Generating summaries...

✅ Done in 8.4s — 4 files processed

── notes.txt ──────────────────────────────
Meeting notes about Q1 product roadmap, focusing on
performance improvements and new integrations.

── todo.txt ────────────────────────────────
Short checklist of 5 items: refactor auth, add tests,
update docs, review PRs, deploy staging.
...

OpenClaw in Your Workflow

Switch between real-world scenarios to see how OpenClaw adapts to any task.

AI-Assisted Code Review

Let OpenClaw automatically review pull requests, suggest improvements, find security issues, and post comments — all driven by GPT-4 or Claude.

Scans diff for bugs and anti-patterns
Suggests idiomatic rewrites
Flags security vulnerabilities

See full example →

$ openclaw run \
  "Review the changes in PR #142 of ./myapp \
   and post a GitHub review comment"

│ ✔ Loading diff (42 changed files)
│ ✔ Analysing with Claude-3.5-sonnet…
│ ⚠ Found 3 potential issues
│ ✔ Posted review to PR #142
✅ Done in 8.2s

Daily Report Automation

Schedule OpenClaw to gather data from multiple sources, analyse it, and send a formatted report to Slack or email — hands-free every morning.

Pulls data from APIs and databases
Generates markdown/HTML reports
Sends via Slack, email, or Discord

See full example →

$ openclaw run \
  "Every weekday at 8am, fetch sales data, \
   generate a Markdown report, and post to Slack"

│ ✔ Scheduled: weekdays 08:00
│ ✔ Data sources: Postgres, REST API
│ ✔ Output: ./reports/2026-03-12.md
│ ✔ Posted to #sales-team
✅ Done in 4.7s

Telegram Chatbot Agent

Deploy OpenClaw as a Telegram bot with persistent memory and tool access. Users can assign tasks via chat, and the agent executes them in the background.

Understands natural language commands
Keeps conversation history
Executes tasks asynchronously

Set up integration →

$ openclaw start --integration telegram \
  --bot-token $TELEGRAM_BOT_TOKEN

🤖 Bot @OpenClawAssist is live!
│ Listening for commands…
│ User: "summarise my emails from today"
│ ✔ Fetching: Gmail API (23 emails)
│ ✔ Summarised with GPT-4o-mini
✅ Summary sent to user

See It in Action

Watch OpenClaw break down a goal and execute it autonomously.

~ openclaw-demo

│ ✔ Goal received: Generate a weekly sales report
│ ✔ Reading: sales_data.csv (1,204 rows)
│ ✔ Analysing data with GPT-4o…
│ ✔ Writing: weekly_report.md
│ ✅ Done in 3.4s — report saved to ./weekly_report.md

Why OpenClaw?

Designed for developers who want control, transparency, and power.

🔒

Privacy First

Run entirely on your hardware with local models. Zero telemetry, zero data sharing. Your prompts are yours.

🔧

Extensible by Design

Write custom tools in Python in minutes. Every component is modular and replaceable via the plugin API.

📋

Transparent Execution

See every action, decision, and tool call the agent makes. Full audit logs and dry-run mode available.

🌐

Any Model, Any Cloud

OpenAI, Anthropic, Google, Mistral, local GGUF — one API interface supports them all through adapters.

OpenClaw vs Other Frameworks

See how OpenClaw compares to popular AI agent alternatives.

Feature	OpenClaw	AutoGPT	CrewAI	LangChain
Local LLM support	✅ Yes	⚠️ Partial	⚠️ Partial	✅ Yes
CLI interface	✅ First-class	⚠️ Limited	❌ No	⚠️ Limited
Telegram / Discord bots	✅ Built-in	❌ No	❌ No	❌ No
Tool plugin API	✅ Simple	✅ Yes	✅ Yes	✅ Yes
Multi-agent workflows	✅ Yes	⚠️ Beta	✅ Core feature	✅ Yes
Dry-run / audit mode	✅ Yes	❌ No	❌ No	❌ No
Zero config for local use	✅ Yes	❌ No	❌ No	❌ No

Loved by Developers

What developers building with OpenClaw have to say.

“OpenClaw replaced a bunch of custom Python scripts for me. I just describe the task and it handles the rest — file processing, API calls, report generation. It just works.”

James S.

Backend Engineer, Berlin

“The local LLM support is a game-changer for our enterprise. No data leaves the server and agents still perform great using Ollama + LLaMA 3. Exactly what we needed.”

Maria K.

DevOps Lead, Warsaw

“I set up a Telegram bot that monitors my CI/CD pipeline and sends me daily summaries. Took 15 minutes with OpenClaw. I’m not going back to writing bots manually.”

Alex T.

Full-Stack Dev, Kyiv