Warp Goes Fully Open-Source: AGPL v3, Agent-First Contribution Model
Warp's complete Rust codebase landed on GitHub under an AGPL v3 license (UI framework: MIT), crossing 37,000–40,000 stars within hours of launch. The headline number isn't the story — the contribution model is: Oz, Warp's cloud agent system, runs multiple AI coding agents in parallel while contributors propose ideas and review output. OpenAI is the founding sponsor powering the GPT-backed agentic workflow; Claude Code, Codex, and Gemini CLI are all first-class plug-in alternatives.
What the Source Actually Says
Two consecutive AlphaSignal issues led with this launch — uncommon for a newsletter that typically features one top repo per issue. The Apr 29 issue grouped it with Zed 1.0 under a single thesis: "raw performance is back in fashion; we're done apologizing for slow tools." The Apr 30 issue sharpened the frame to a structural claim: "the dev tooling layer is being rebuilt around agents, not humans."
The licensing structure carries concrete product implications. AGPL v3 means any publicly distributed fork must remain open-source — a contribution obligation that may deter enterprise forks while accelerating OSS iteration. The MIT UI layer carries no such constraint. Warp's Oz system orchestrates multiple coding agents in parallel; a new /feedback skill lets contributors open GitHub issues from within the terminal without switching context. CONTRIBUTING.md is the designated entry point for the agent-driven workflow.
The Apr 30 issue made the structural pair explicit: Cursor simultaneously released an SDK (npm install @cursor/sdk) that breaks its coding agents out of the editor into CI pipelines, bug-to-PR loops, and internal tools. Both bets converge on the same thesis — agents become infrastructure, not editor features.
Strategic Take
The multi-LLM plug-in stance is the signal worth watching: Claude Code, Codex, and Gemini CLI are all listed as first-class agents inside Warp. A terminal that ships with explicit multi-model support is betting on commoditized inference, not vendor lock-in. The AGPL v3 internals are now auditable for the first time — teams evaluating agent runtimes have a production-grade Rust codebase to study.
