Community fork of Warp terminal that adds BYOP (bring your own provider) via OpenAI-compatible endpoints, minijinja prompt templates, and local-only credential storage.
Key Takeaways
Supports OpenAI, Anthropic, DeepSeek, Qwen, Groq, and local Ollama/LM Studio via any OpenAI Chat Completions streaming endpoint; configured in ~/.config/openwarp.toml.
System prompts use minijinja templates with context vars like {{ cwd }}, {{ locale }}, and {{ user.role }} for dynamic rendering.
API keys never leave the device; no telemetry, no cloud upload, direct connection to your specified Base URL.
Tracks Warp upstream (AGPL/MIT dual license), preserving blocks, workflows, AI commands, and keymaps.
Early development, no stable release yet; multilingual UI ships with zh/en/ja/es.
Hacker News Comment Review
Warp founder confirmed BYOM is already planned natively in Warp, with an active GitHub discussion, undercutting the core use case of this fork.
A commenter who installed from source reported still hitting a $20/month signup gate, contradicting the BYOP premise; the README omits this friction entirely.
Broader consensus: many users want Warp’s original terminal UX (block history, IDE-like prompt) stripped of AI, not more AI layering; OpenWarp does not address that.
Notable Comments
@mark_l_watson: installed from source and still hit the $20/month paywall before custom providers work – not disclosed in README.
@tyleo: project website is severely broken on Safari/iOS with overflow, reflow animations, and non-English text mixing – trust signal problem for a security-sensitive tool.
