Warp Code

Warp Code is the agentic coding surface of Warp's broader development environment, designed to let developers write rich prompts, let agents scope and plan work, review output inline, and ship to production — all without leaving the terminal. The platform is built by Warp, whose terminal client is open-source under AGPL-3.0 and has accumulated over 63,000 GitHub stars. Warp says the platform is used by over 800,000 developers and thousands of engineering teams.

What It Is

Warp Code is the coding-focused layer of Warp's Agentic Development Environment (ADE). It combines a modern terminal, a native file editor, codebase indexing, and AI agent orchestration into a single surface. Rather than switching between a code editor, a terminal, and a separate AI chat tool, developers stay in one interface: prompt an agent, review its plan, edit inline, and push to a PR. The underlying terminal client is written in Rust and is open-source; the cloud agent orchestration layer (called Oz) is a separate hosted platform.

How the Workflow Works

The core loop in Warp Code follows four stages the product page labels Prompt → Plan → Refine → Ship:

• Prompt: Write rich prompts with file references, images, and voice-to-text (via Wispr Flow integration).
• Plan: Agents scope the work, surface tradeoffs, and produce an editable plan before writing code.
• Refine: Review agent output inline, leave comments, and re-prompt with one click without breaking flow.
• Ship: Stay in the terminal from prompt to PR; open, edit, and save files in Warp's native file editor.

Multi-Model and Multi-Repo Architecture

Warp Code is multi-model by default. It provides instant access to frontier models from OpenAI, Anthropic, Google, and xAI, plus open-source models such as GLM 5.2 hosted on Fireworks AI in the US. A mixed-model routing mode called Auto selects the best model per step. The Oz agent layer supports multi-repo awareness, letting agents coordinate cross-team changes such as client-server contract updates or internal documentation across repositories. Oz agents also have Full Terminal Use (running interactive terminal commands) and Computer Use (verifying changes end-to-end).

Privacy and Security Controls

Warp Code includes a layered set of controls aimed at enterprise and security-conscious teams:

• Tunable agent autonomy from step-by-step approval to full autonomy, configurable at individual and org level
• Zero Data Retention agreements with supported LLM providers (Anthropic, OpenAI, Google, xAI, Fireworks AI)
• Secret redaction with custom regexes so sensitive values never reach a model
• Password manager integrations with 1Password and LastPass
• SAML-based SSO and role-based access control on higher-tier plans
• SOC 2 Type II attestation, with a public Trust Center
• One-click analytics and telemetry opt-out; real-time Network Log for transparency

MCP and Integrations

Warp Code supports the Model Context Protocol (MCP), allowing users to connect external services directly into the agent context. The product page lists Linear, Figma, Slack, and Sentry as examples of MCP-connected services. Codebase indexing supports large professional codebases with real-time indexing, and users can supply Rules, images, docs, and URLs to tailor agent responses.

Update: Open-Source Launch and Oz Multi-Harness Orchestration

Warp announced that Warp Terminal is now open-source, with the client codebase published under AGPL-3.0 (UI framework crates under MIT) on GitHub. The latest stable release is v0.2026.06.03.09.49.stable_00, published June 3, 2026. Separately, Warp launched multi-harness orchestration in Oz, enabling users to orchestrate Claude Code, Codex, and the Warp Agent together in cloud environments — mixing and matching agent harnesses across workflows. OpenAI is noted as the founding sponsor of the open-source repository, and the agentic management workflows in the repo are powered by GPT models according to the GitHub README.