Frequently Asked Questions

OpenClaw is an open-source AI Agent framework that executes real tasks on your computer—reading files, running commands, modifying code—not just chat. Think of it as a junior engineer with sudo privileges. If you wouldn't trust an intern with root access to this folder, don't trust the agent.

Yes. "Clawdbot" was the previous name. The project evolved rapidly into OpenClaw. ⚠️ **Warning:** Most "Clawdbot" tutorials online are now **outdated or unsafe** because they reference deprecated dependencies and Windows pathing issues that no longer work. This site tracks the **verified current reality**—not ghost maps from abandoned rebrands.

ChatGPT answers questions and gives advice. OpenClaw reads files, runs commands, modifies code, and executes workflows. ChatGPT is a consultant. OpenClaw is the intern who actually does the work.

No. OpenClaw is a scheduler that connects to external LLM providers (OpenAI, Claude, Ollama, etc.). It doesn't sell models, just makes models 'work'.

If you use Ollama locally, yes—the model runs on your machine. To verify: block outbound traffic (except localhost) via Little Snitch or `ufw`. If OpenClaw still talks to your local Ollama, it's running locally. If it hangs, check your `base_url`. If you use API providers (DeepSeek, OpenAI, Anthropic), your prompts go to their servers.

Yes. Set `LLM_PROVIDER=openai` and `BASE_URL=https://api.deepseek.com`. Example `.env`: ```bash LLM_PROVIDER="openai" LLM_BASE_URL="https://api.deepseek.com" LLM_API_KEY="sk-your-key-here" LLM_MODEL="deepseek-reasoner" ``` 👉 See our DeepSeek Config Guide for OpenClaw for full setup.

Yes. Use `provider: ollama`. Example setup: ```bash # Install Ollama & pull model curl -fsSL https://ollama.com/install.sh | sh ollama run deepseek-r1:8b # Configure OpenClaw LLM_PROVIDER="ollama" LLM_BASE_URL="http://localhost:11434/v1" LLM_MODEL="deepseek-r1:8b" ``` ⚠️ Requires heavy hardware. See Hardware Reality Check.

Ollama runs the DeepSeek model (loads it into VRAM, handles inference). OpenClaw tells it what to do (reads files, runs commands, executes workflows). If Ollama is down, OpenClaw is useless. If OpenClaw isn't running, Ollama is just a chatbot. You need both.

Basic use: no coding needed, but basic logic helps. Advanced use: knowing command line and project structure helps. It's not 'zero barrier', but 'low barrier, high ceiling'. If you don't know what `chmod +x` means, you're going to have a bad time.

Mac: most friendly. Linux/Server: first choice for production. Windows: usually via WSL2 (strongly recommended). If you see `Error: connect ECONNREFUSED 127.0.0.1:11434`, that's a WSL2 networking issue. Mac users suffer slowly (3.2 tokens/sec). Windows users suffer dramatically (WSL2 drama). Linux users just suffer.

Yes. It can run long-term, retry on failure, save intermediate state, and stop by rules—which is why it's called an autonomous agent. That's also why it's a security risk: it doesn't know when to quit.

DeepSeek R1 wraps responses in `` tags before the actual JSON. OpenClaw's JSON parser fails. Error: `SyntaxError: Unexpected token <` (the model is 'thinking' out loud) 👉 Fix it here: JSON Parsing Fix Guide.

OpenClaw has powerful capabilities but requires strict guardrails (containerization, permission limits, command blocking) to be safe. If you wouldn't trust a junior intern with root access to this folder, don't trust the agent. Real incidents I've stopped: - Agent tried to `rm -rf .` to "clean build artifacts" - Agent attempted `curl unknown.sh | bash` because it needed a tool Mitigation strategies: - Run in Docker container with read-only filesystem - Use dedicated device (Mac Mini, cheap server) - Block dangerous commands (rm, format, dd, etc.) - Review every execution log 👉 Read the full autopsy: CVE-2026-25253 Analysis.

If you give too many permissions, yes. OpenClaw's capabilities equal the permissions you give. Correct approach: read-only by default, specify writable directories, block dangerous commands. 'Going rogue' is just a fancy way of saying 'it did exactly what you told it to do, not what you meant'.

Yes, but only if you have strict guardrails. Otherwise, expect to wake up at 3 AM. You need: - Exact understanding of what the agent can and cannot do - Every workflow tested in staging - Permission isolation (read-only by default) - Logging and rollback mechanisms - Human review of every action If you're missing any of these, you're not ready for production.