Suggestion for Adding Support for Custom AI Model Configuration

[LeoYoung-code]

Pre-submit Checks

• I have searched Warp feature requests and there are no duplicates
• I have searched Warp docs and my feature is not there

Describe the solution you'd like?

Dear Warp Team,

I hope this message finds you well. First of all, I would like to express my gratitude for developing such an innovative and user-friendly
terminal tool. Warp has significantly enhanced my workflow, and I truly appreciate the effort and dedication that has gone into creating
this tool.

I am writing to kindly suggest an enhancement that could further improve the flexibility and customizability of Warp. Currently, Warp does
not support the configuration of custom AI models. While the default AI model is excellent, there may be scenarios where users would
benefit from the ability to configure their preferred AI model.

Allowing users to customize the AI model could enable a broader range of use cases, tailored to individual preferences and specific
requirements. This feature could also encourage more advanced users to experiment with different models, potentially leading to more
diverse and creative uses of Warp.

I understand that implementing this feature may require significant development effort and consideration of various technical aspects.
However, I believe that this enhancement could add significant value to the tool and cater to a wider audience.

Thank you very much for taking the time to consider my suggestion. I am confident that the Warp team will continue to innovate and improve
the tool, and I look forward to seeing the future developments.

Best regards,
leo

Is your feature request related to a problem? Please describe.

No response

Additional context

No response

Operating system (OS)

macOS

How important is this feature to you?

1 (Not too important)

Warp Internal (ignore) - linear-label:39cc6478-1249-4ee7-950b-c428edfeecd1

None

[c0des1ayr]

+1, but preferably, there should also be the option to use an external model (Gemini, Llama, etc.), whether it is local or remote.

[zerx-lab]

+1, but preferably, there should also be the option to use an external model (Gemini, Llama, etc.), whether it is local or remote.

[koiker]

To make this feature more flexible. Warp should be able to accept any model based on OpenAI API standard. Allowing users to input the endpoint address and based on this endpoint and, using the API standard from OpenAI, interact with the LLM model.
This will make the Warp terminal compatible with any LLM implementation. Be this vLLM, Dynamo, Ollama or any other inference service local or remote.

[Temjinck]

This is a must-have feature if you want Warp to be used much more widely.

[jbaenaxd]

This is precisely the reason why I canceled my subscription after just one month. The paid plan is incredibly expensive for such a limited number of requests, especially considering the availability of numerous code-cli applications for the terminal, such as OpenAI Codex, Gemini CLI, and Claude Code. These alternatives offer similar functionalities at significantly lower costs.

Furthermore, their decision to exclude external AIs stems from their own greed.

[c0des1ayr]

Yes, it seems that custom AI models are an ""enterprise"" feature. Absolute garbage, better to move on to Wave Terminal at this point.

[myleshyson]

I just realized that using your own ai models is an enterprise feature. Cancelled the subscription 5 minutes after I purchased it. Maybe i'm dumb and that's a good business model. Just seems like you'd get a lot more people willing to pay $20 a month to use your terminal AND use any openapi compatible api. At least that was the case for me.

[ThePlenkov]

It would be really nice! As a user i use terminal sometimes with sensitive information - for example sharing tokens and other stuff - for that reason i intentionally want to use a self-hosted LLM for a terminal experience ( i don't really need opus to suggest how to grep ). As terminal I see warp one of the best tools. But since is using own API - it becomes restricted for use in most of enterprises by default..

[ThePlenkov]

Just FYI i found that wave terminal has a similar feature already in-place

{
  "conn:askbeforewshinstall": false,
  "waveai:defaultmode": "litellm-gemini-flash-3.1-lite",
  "waveai:showcloudmodes": false
}