Opencode - Always "error=Model tried to call unavailable tool"

[Flagelmann]

Question

Hello,

Just a simple question: how to make opencode finally be able to analyze my code base without copy and paste its code manually as any other chatbot?

On the paper this software should be able to interact with my code directly, ask questions about functions, reworking stuff, etc. but I'm spending more time fixing and making it "working" than using it properly.

Last question is about these "tools", do they works on Windows platforms or not?

As I'm just getting errors "error=Model tried to call unavailable tool.....etc...." for any action I try to ask in Opencode.

I'm using qwen3.5 successfully in other AI tools, dunno why with this everything is so difficult to setup....

I think I will drop it anyway, but I'm just curious how these tools should "work" in it.

Thanks

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• LM Studio (local) model is not executing tool correctly #20836: LM Studio (local) model (Qwen) not executing tools correctly on Windows — very similar symptoms
• Thinking but no answer or action when using local models #20906: Qwen model via Ollama only thinks but doesn't use tools to act on the codebase
• Non-Claude models fail to call task tool due to required subagent_type parameter #20804: Non-Claude models (including Qwen) fail to call tools due to parameter name mismatches
• Agent loop stops after first LLM call when using @ai-sdk/openai-compatible with LiteLLM (finish_reason: stop instead of tool_calls) #20719: Agent loop stops after first LLM call when using OpenAI-compatible providers (finish_reason: stop instead of tool_calls)

The "Model tried to call unavailable tool" error with Qwen models on Windows is a known pattern with local/OpenAI-compatible providers. These existing issues may have workarounds or context that could help.

[ptrxyz]

I am also running into this basically everytime i am using Qwen. is there a way to find out what tool the llm is trying to call? issue seems present in plan as well as in build mode.

the logs do not really reveal a lot:

invalid [error=Model tried to call unavailable tool 'unknown'. Available tools: bash, ... (many more) ..., tool=unknown]

Can we workaround this by putting something smart in the AGENT.md file to limit the tools that the agent can use?

[sk-surya]

same issue here

[jonas-klesen]

Happens constantly with Qwen 3.7Max. FWIW, I am on Linux

[jensenojs]

glm5.1 here

[mjzavar]

i had the same problem , solved it by increasing the context length in ollama

[BernhardGruen]

Please have a look at this PR: #33504
I am quite sure it will fix your problems too.

Also: could you please provide information which provider you are using.