← All guides

OpenClaw + Qwen: Verified Setup Paths for OAuth and Ollama

If you searched for OpenClaw Qwen, there are two official paths: use the built-in Qwen OAuth provider, or run a local Qwen model through Ollama. Here is the shortest version that stays inside the official docs.

The Quick Answer

OpenClaw officially supports Qwen OAuth through the qwen-portal provider and also works with local Qwen models through Ollama.

Watch: Qwen 3.6 + Ollama End-to-End Walkthrough

Running a local open-weights model (Qwen 3.6) through an agent runtime looks the same whether you’re pairing it with OpenClaw, Qwen Code, or any other Claude Code alternative. This 16-minute video walks through the full setup — installing Ollama, pulling the model, wiring the agent, and three real coding demos against Claude Code.

Apply the same Ollama wiring below for OpenClaw — the agent runtime is the only piece that differs. Want to see Qwen 3.6 27B pushed harder on real code? Watch the full uncut local coding session where it goes head-to-head with Claude Code, or read the RTX 3090 benchmark breakdown.

Option 1: Use the Official Qwen OAuth Provider

# 1. Enable the Qwen auth plugin
openclaw plugins enable qwen-portal-auth

# 2. Restart the gateway
openclaw gateway restart

# 3. Start Qwen login and set it as default
openclaw models auth login --provider qwen-portal --set-default

# 4. Switch to the coder model explicitly
openclaw models set qwen-portal/coder-model

The flow uses Qwen’s device-code OAuth with a free-tier limit of 2,000 requests per day.

Option 2: Run a Local Qwen Model Through Ollama

# 1. Pull a Qwen model into Ollama
ollama pull qwen2.5-coder:32b

# 2. Let OpenClaw talk to Ollama
export OLLAMA_API_KEY="ollama-local"

# 3. Verify the model is visible
ollama list
openclaw models list

# 4. Set it as the default if needed
openclaw models set ollama/qwen2.5-coder:32b
🖥️ RECOMMENDED HARDWARE FOR QWEN

Qwen runs best with RAM headroom. A 24 GB Mac is the practical minimum; 48 GB+ runs the bigger Qwen variants comfortably.

The Most Important Ollama Caveat

Do not point OpenClaw at the OpenAI-compatible /v1 endpoint. The official Ollama page warns that native tool calling is the reliable path.

# Good: native Ollama API
http://ollama-host:11434

# Risky for tool calling
http://ollama-host:11434/v1

How to Verify Your Qwen Setup

openclaw models list
openclaw status
openclaw doctor
openclaw dashboard

If you are still setting up OpenClaw itself, start with How to Install OpenClaw. If Qwen is configured but replies are blank or tools fail, jump to the troubleshooting guide.

Need OpenClaw fixed live?

Remote rescue sessions for gateway, auth, tunnel, VPS, and model access problems.

See Rescue Session

Read next

Best Local LLM Reddit Picks for OpenClaw (2026)
A Reddit-intent answer for best local LLM reddit searches: practical OpenClaw picks by RAM, VRAM, context needs, and tool-call reliability.
Best OpenClaw Model Reddit Users Recommend (2026)
The Reddit-intent answer for best OpenClaw model searches: Qwen for daily local work, gpt-oss for tool-call reliability, Scout for long context, and how to choose by RAM or GPU.
Ollama Local LLM Reddit Picks for OpenClaw (2026)
A Reddit-intent Ollama answer for OpenClaw: which local LLM to pull first, when to use Qwen, when to use gpt-oss, and why RAM/VRAM headroom matters more than model size.