"model does not support tools"
This error means your Ollama model was never trained for function calling, so it can chat but cannot call tools. Switch to a tool-capable model like qwen3-coder:30b, hermes3, or mistral:7b, and make sure your baseUrl uses the native API with no /v1 suffix, since /v1 streaming drops tool-call chunks. Then restart the gateway.
Likely cause
Many small or base models (phi3:mini, qwen2.5:3b) were never trained for function calling, so the runtime refuses tool calls. A second cause: even a tool-capable model fails silently if your baseUrl ends in /v1, because the /v1 streaming path drops tool-call chunks.
The fix
- 1 Switch to a tool-capable model such as qwen3-coder:30b, llama3.1:8b, hermes3 or mistral:7b.
- 2 Pull it with Ollama.
- 3 Make sure your Ollama baseUrl has no /v1 suffix (use the native http://127.0.0.1:11434).
- 4 Point your OpenClaw config at the new model name and restart the gateway.
ollama pull qwen3-coder:30bHit a different error?
Paste any agent error and get the cause and fix in seconds.
Frequently asked questions
Is this a bug in OpenClaw or my model?
Neither is broken. The model genuinely was not trained for function calling, so it cannot generate tool calls. Switching to a tool-capable model fixes it.
I switched models and tools still fail silently. Why?
Your baseUrl probably ends in /v1. Ollama's /v1 streaming path drops tool-call delta chunks, so even a tool-capable model fails. Use the native http://127.0.0.1:11434 instead.
How do I know if a model supports tools before pulling it?
Check the model's page on ollama.com for 'tools' or 'function calling' in its capabilities. If it isn't listed, the model won't work for agent tasks.
Stop firefighting agent errors
Decoding errors one at a time is the manual version of what BetterClaw automates. Run your OpenClaw agents hosted with managed models, retries and config validation built in.
$19/month per agent · BYOK · 7-day money-back guarantee