Why Does the Tools Agent in n8n Only Work Reliably with OpenAI's Model?

It sounds like you're encountering some of the edge cases we’ve seen when using non–OpenAI Chat Models with the Tools Agent. The documentation recommends using the OpenAI Chat Model for the best and most consistent experience since it has been extensively tested with tool-calling capabilities.

Other models (such as Groq, Mistral Cloud, Anthropic, and Azure OpenAI Chat Model) are supported, but their support may depend on how well they handle the formatting and tool calling conventions that our Tools Agent expects. Inconsistent behavior or errors with these models might not be due to a misconfiguration on your end but rather differences in how these models interpret the instructions for tool use.

If you’re running into repeated issues with non–OpenAI models, consider the following steps:

• Ensure your node configuration, especially the prompt and output format settings, strictly follow the recommended formats.
• Verify that the Chat Model sub-node is correctly connected and configured.
• Check if there are any recent updates or compatibility notes in the n8n documentation for the specific model you’re trying to use.

In short, many users have reported that the OpenAI Chat Model offers the most reliable experience with the Tools Agent node, so continuing with it might be the best option unless you need to experiment with alternatives.