Ollama + OpenClaw Quick Start | macOS

<img src="/images/dock_head_s.png" alt="Duck Editor" width="24" style="vertical-align: middle;"> Duck Editor, straight to the point: Ollama + OpenClaw needs no API key setup at all. Use the free cloud quota from your Ollama account, pick a model, and start chatting. The whole flow looks like this:

Install Ollama → ollama launch openclaw → Pick a cloud model → Start chatting

---

Step 1: Install Ollama

1. Go to ollama.com/download
2. Click “Download for macOS”
3. Open the downloaded .dmg and drag Ollama into your Applications folder
4. Open Ollama

🚨 macOS security prompt: If you see “Ollama can’t be opened because it is from an unidentified developer,” go to “System Settings → Privacy & Security,” scroll to the bottom, and click “Open Anyway.”

Confirm the install succeeded:

ollama --version

If you see a version number, you’re good: ollama version 0.6.x

---

Step 2: Launch OpenClaw With One Command

Open Terminal:

Type:

ollama launch openclaw

The first run installs automatically. When it asks whether to install via npm, press Enter to confirm:

After installation, pick a model. Just choose “Cloud” — there’s no need to download a model to your machine.

<img src="/images/dock_head_s.png" alt="Duck Editor" width="24" style="vertical-align: middle;"> Recommended cloud models (use Ollama’s free quota, no download):

Model	Strengths
minimax-m2.5	🥇 Tested pick: best tool-calling ability, excellent on Agent tasks
kimi-k2.5	1T parameters, strong conversation quality
glm-4.7	General-purpose, stable and reliable

One note though: these models are fine for playing around, but if you want to roll them into a real workflow, it’s still better to switch to mainstream models like Claude or GPT-4o — there’s a clear gap in stability and tool-use reliability.

macOS may show a security prompt (Node.js is OpenClaw’s runtime); just click “Allow”:

When the ready screen appears, you’ve launched it successfully!

---

Step 3: Your First Conversation — Name Your Lobster 🦞

Try typing:

Hi! I'm the owner here. Please introduce yourself in English — who you are and what you can do.
Also, pick a distinctive name for yourself!

Congratulations 🎉 Your lobster is officially up and running!

<img src="/images/dock_head_s.png" alt="Duck Editor" width="24" style="vertical-align: middle;"> If the model replies in a language you don’t want, just tell it: “Please reply in English,” and it’ll switch.

---

About the Free Quota

Ollama’s cloud models come with a free quota (official details). To check your usage:

1. Go to ollama.com/settings and sign in with your Google account

Out of quota? You can switch to the Gemini Flash cloud API, or download an open-source model locally (see the section below).

---

FAQ

🚨 OpenClaw Can’t Connect to Ollama

# Confirm Ollama is running
ollama ps

# Confirm the API is healthy
curl http://localhost:11434/api/tags

Make sure the Ollama icon is in your menu bar. If it isn’t, reopen the Ollama app.

🚨 The Model Responds Slowly

Cloud models aren’t affected by your machine’s specs; speed differences only appear once you switch to a local model.

🚨 I Want to Switch Models

Re-run ollama launch openclaw, or switch from the settings page:

ollama launch openclaw --config

---

Bonus: Download an Open-Source Model Locally (Optional)

Want to go fully offline, save cloud quota, or just curious? Here’s the quick version.

Check your specs (local models need at least 8 GB RAM):

Your hardware	Recommended model	Expected speed
8GB RAM, no GPU	llama3.2:3b	5–10 tokens/sec
16GB RAM, Apple M1	qwen2.5:7b	15–25 tokens/sec
32GB+ RAM, Apple M series	qwen2.5:14b	30+ tokens/sec

# Download (recommended for Chinese users)
ollama pull qwen2.5:7b

# Test
ollama run qwen2.5:7b

Apple Silicon (M1/M2/M3/M4) enables GPU acceleration automatically — no extra setup needed.

---

Next Steps

• ⚙️ Model Configuration and Switching
• 🧩 Build Your First Skill
• 📱 Connect Telegram

Questions? Join the homepage discussion and let’s talk!