Warp, the AI-powered coding tool, has unveiled a strategy to make coding agents more understandable — and the approach bears a strong resemblance to pair programming.
Today, the company is launching Warp Code: a new suite of features crafted to give users greater oversight of command-line-based coding agents. These features include more comprehensive difference tracking and a clearer view of the coding agent’s activities, addressing a key pain point for developers relying on AI tools.
Founder Zach Lloyd explains the motivation behind the update: “With many other command-line tools, you’re essentially crossing your fingers and hoping that the output from the agent is something you can actually merge into your project.” The new features, he adds, are designed to “create a far tighter feedback loop for this agent-driven style of coding.”
In practical use, this means developers can monitor exactly what the agent is doing and pose questions in real time. “As the agent writes code, you’ll be able to see every small diff it generates,” Lloyd notes. “And you’ll have a simple way to comment on those diffs and adjust the agent’s direction as it works.”
The interface will feel familiar to existing Warp users: there’s a section at the bottom for issuing direct instructions to the agent, a window for viewing the agent’s responses, and a side panel that shows the agent’s changes step by step. Like code-focused tools such as Cursor, users can edit code manually if they prefer. Additionally, they can highlight specific lines to provide context for requests or questions. One of the most standout features? Warp’s compiler automatically troubleshoots errors that arise during code compilation, saving developers time on debugging.
Lloyd emphasizes the core goal: “This is about ensuring you understand the code the agent produces — and making sure you can edit and review it with ease, just like you would with code written by a human collaborator.”
This launch marks a fresh approach to the increasingly competitive landscape of AI-driven programming. Warp faces competition from a range of players: fully no-code tools like Lovable, AI-enhanced code editors such as Cursor and Windsurf, and even foundation model companies that offer their own command-line tools (e.g., Anthropic’s Claude Code and OpenAI’s Codex) — all while Warp itself leverages these companies’ models to power its own product.
While Warp currently has 600,000 active users and counting (making it a relatively small participant in the AI coding race), its growth is accelerating rapidly. Lloyd reveals that the company adds $1 million in Annual Recurring Revenue (ARR) every 10 days — a sign that many developers are willing to pay for a more refined “vibe coding” experience.