zerx-lab · 2026-04-29 · major
OpenWarp — Community Fork of Warp Opens Up Bring-Your-Own AI Provider
Days after Warp open-sourced its terminal client, this community fork strips out the cloud-tied AI layer so any OpenAI-compatible endpoint — OpenAI, Anthropic, DeepSeek, Ollama — works locally with no relay.
Plug any OpenAI-compatible model into Warp's terminal — credentials stay local, no cloud relay.
Key specs
| License | AGPL-3.0 / MIT (dual) |
|---|---|
| GitHub stars | 193 |
| Language | Rust |
What is it?
OpenWarp is a community fork of warpdotdev/warp that opens the AI provider layer Warp's mainline still routes through its cloud account. It keeps Warp's blocks, AI commands, workflows, and keymaps intact and continuously merges upstream code.
How does it work?
It points the AI plumbing at any OpenAI-compatible Chat Completions endpoint — OpenAI, Anthropic gateways, DeepSeek, Qwen, local Ollama — with credentials stored only in the local config file. A minijinja template engine renders dynamic system prompts from the current directory, locale, and role. Cloud-account refresh, persistent user storage, plan sync, and passive-suggestion HTTP are all disabled, so requests only go to providers the user configures.
Why does it matter?
Warp open-sourced its client on April 22 with OpenAI as founding sponsor, but the AI layer still phoned home. OpenWarp answers the obvious next question — 'can I run this against my own model?' — three days after the source dropped. Warp's own founder confirmed in the HN thread they'll add bring-your-own-model upstream, so this fork is forcing the timeline.
Who is it for?
developers who want Warp's terminal UX with local or self-hosted models
Try it
git clone https://github.com/zerx-lab/warp