Introduction
Warp, the modern Rust-based terminal and agentic development environment, has taken a bold step by open-sourcing its client code. But this isn't your typical open-source project where you fork, code, and submit a pull request. Warp has designed a contribution model that leans heavily on AI coding agents—specifically its own Oz platform—to speed up development. The idea is that human contributors focus on decision-making and review, while AI agents handle the implementation. If you're excited to get involved, this guide will walk you through exactly how to contribute using Warp's new workflow. By the end, you'll be ready to suggest features, verify behavior, and let AI write the code—all within the Warp ecosystem.
What You Need
- A GitHub account
- Familiarity with the Warp terminal (installed on Linux, Windows, or macOS)
- Access to Oz, Warp's cloud agent orchestration platform (free tier available)
- An OpenAI API key or other supported model access (Warp recommends Oz for best integration)
- Basic understanding of terminal features and software testing
- A text editor or IDE for spec work and reviews
Step-by-Step Guide
Step 1: Clone the Repository and Explore the License
Head over to the official Warp repository on GitHub at github.com/warpdotdev/warp. Before you dive in, note the split licensing: the UI framework (warpui_core and warpui crates) is under the MIT license, while the rest of the codebase uses AGPLv3. Make sure you're comfortable with these terms, especially the AGPLv3 clauses if you plan to distribute modified versions. Fork the repo to your own account and clone it locally.
Step 2: Understand the Agent-Driven Workflow
Warp's open-source model is unique: the main bottleneck isn't writing code—it's human effort in defining features and verifying behavior. As a contributor, your primary role is to create specs and review AI-generated code. Oz, Warp's cloud agent orchestration platform announced earlier this year, lets you run multiple coding agents in parallel with full visibility. The agents, powered by GPT models (OpenAI is the founding sponsor), write the actual code based on your specs. Other coding agents like Claude Code or Codex are welcome, but Warp recommends Oz because it already has the right context and verification checks baked in.
Step 3: Set Up Oz and Configure Your Agent
If you haven't already, sign up for Oz (warp.dev/oz). Once you're in, create a new agent session. Warp suggests using the 'auto (open)' routing option that selects the best open-source model for the task. You can also choose from additional models now supported: Kimi, MiniMax, and Qwen. Configure a settings file for programmatic control—this makes your setup portable across devices and repeatable for different tasks. Ensure your agent has access to the forked Warp repository.
Step 4: Write a Spec for a New Feature or Fix
Think of a change you'd like to see in Warp—maybe a new block command, a UI tweak, or better integration with a tool. Write a clear, concise spec. This includes what the feature should do, how it should behave, and any acceptance criteria. Keep it focused; the AI will implement it, so your spec acts as the blueprint. Share your spec in the community discussion or directly as a GitHub issue tagged with 'agent-spec'. This is where human-led contribution shines.
Step 5: Let the Agent Write the Code and Verify
Feed your spec to your Oz agent (or your preferred coding agent). The agent will generate code following Warp's rules and verification processes. Monitor the agent's progress in the Oz dashboard—you have full visibility into each step. Once the code is ready, test it manually or using automated tests. Warp is confident that human contributors are in a good position to get features right by verifying the AI's output. If something is off, iterate on the spec or ask the agent to refine.
Step 6: Submit Your Contribution for Review
After verification, commit the AI-generated code to your fork and create a pull request back to the main Warp repository. In the PR description, link to the original spec issue and mention that the code was generated by an agent via Oz. Warp's team and the community will review your submission. Remember, this workflow is designed so that human effort is spent on high-level decisions and quality checks, not on writing every line of code.
Conclusion & Tips
- Start small: Begin with minor bug fixes or simple UI changes to understand the agent workflow. Warp's repository has a 'good first issue' label for newcomers.
- Leverage Oz's visibility: Use the Oz dashboard to see exactly what the agent is doing—this helps you learn the codebase and catch errors early.
- Keep specs tight: Overly broad specs confuse agents. Break features into bite-sized, testable pieces.
- Test thoroughly: Even though agents follow verification rules, always run the tests yourself. Warp values quality over speed.
- Engage with the community: Warp's open-source shift is new—share your experience and suggestions in GitHub discussions. Your feedback shapes the project's future.
- Stay updated: Warp continues to expand open-source model support and tooling. Watch the repository for changes to contribution guidelines.
Warp's decision to open-source and embrace agentic contributions marks a shift in how open-source projects can evolve. By focusing human talent on ideas and review, and letting AI handle the grunt work, the project accelerates development while keeping quality high. Whether you're a seasoned terminal user or a curious developer, this new model invites you to participate in a fresh way. Dive in, write a spec, and let the agents do the coding—under your watchful eye.