registry.ollama.ai/library/… does not support tools

Likely cause

Cause 1: the model's Ollama build has no tool-calling template, so Ollama refuses any request that includes tools. That covers older and specialised models such as llama3, gemma3, dolphin3, phi4 and vision models like llava, even when the underlying weights can follow instructions. Cause 2: 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. 1 Switch to a tool-capable model: qwen3.5:9b on a 16GB machine, qwen3.8:27b with 24GB or more, or gemma4:12b, hermes3:8b, llama3.1:8b or mistral:7b.
2. 2 Pull it with Ollama and confirm the tools capability with ollama show.
3. 3 Make sure your Ollama baseUrl has no /v1 suffix (use the native http://127.0.0.1:11434).
4. 4 Point your app at the new model name (see the per-app fixes below) and restart it.

Does your model support tools?

Models that support tools include qwen3.8, qwen3.5, gemma4, hermes3, llama3.1 and mistral. Models that don't include llama3, gemma3, gemma2, dolphin3, phi4 and llava. Find the model name from your error below.

Checked against each model's capability tags on ollama.com in September 2026. Tags change as models are updated, so confirm locally with ollama show.

Cause 3: your custom Modelfile dropped the tool template

Still getting the error on a model from the "Yes" list? If you built a custom variant with ollama create (to raise the context window or set a system prompt), check its template. Ollama only enables tools when the template references .Tools. Importing a raw GGUF, or writing your own TEMPLATE line, can leave that block out, so the custom model loses tools even though the base model has them.

# 1. Check the custom model's template for a .Tools block
ollama show mymodel --template | grep -i tools

# 2. If nothing prints, start from the base model's full Modelfile
ollama show qwen3.5:9b --modelfile > Modelfile

# 3. Edit only the PARAMETER / SYSTEM lines you need, e.g.
#    PARAMETER num_ctx 32768

# 4. Recreate the model
ollama create mymodel -f Modelfile

Copy the template from the same model family you are customising. Templates are model-specific, so a template from a different model will produce broken tool calls.

Fix for your specific app

The error always comes from Ollama, but where you change the model depends on which app sent the request.

OpenClaw

Set a tool-capable model as your primary in the gateway config, or switch in chat with /model, then restart the gateway.

primary: "ollama/qwen3.5:9b"

Hermes Agent

Set the default model, or run hermes model to pick one interactively.

hermes config set model.default qwen3.5:9b

Open WebUI

Pick a tool-capable model in the model selector at the top of the chat. To change the default for everyone, use Admin Panel → Settings → Models.

n8n

Open the Ollama Chat Model sub-node attached to your AI Agent node and change the Model field to one from the "Yes" rows above.

LangChain / LlamaIndex

Change the model argument where you construct the client, then bind your tools as before.

llm = ChatOllama(model="qwen3.5:9b")

Any other app

Find where the model name is configured and swap it for a model from the "Yes" rows above.