open-warpBring your own AI backend to Warp.

A local open-source adapter that routes Warp AI requests to OpenAI-compatible providers such as DeepSeek, OpenAI, Ollama, OpenRouter, LM Studio, and vLLM.

Get Started

View on GitHub

Local by design

WarpLocal routes AI requests to a Go helper running on your machine instead of the official Warp cloud backend.

Bring your own provider

Configure any OpenAI-compatible endpoint, API key, and model from the WarpLocal settings menu.

Coding-agent workflow

Supports core tools for reading files, searching code, running shell commands, applying file diffs, and collecting command output.

Works beside official Warp

The patched WarpLocal app uses a separate local channel and can coexist with the official Warp app.

What is open-warp?

Warp's client is open source, but its AI backend is still controlled by the official cloud service. open-warp fills that gap with a local adapter server.

The project ships two parts:

WarpLocal.app: a patched Warp app bundle that sends AI traffic to 127.0.0.1:18888.
warp-local-adapter: a Go backend that translates Warp's protobuf stream into OpenAI-compatible chat completions.

Current status

open-warp is usable for the core coding loop: ask, inspect files, search, run commands, apply diffs, and continue with tool results. It does not yet aim for complete Warp backend parity.

If this project helps you, please star sasuke39/open-warp. That signal helps guide future tool support and release work.