Warp Terminal Goes Open Source: 37K Stars in Days

Warp terminal went open source in May 2026, releasing its client codebase on GitHub under AGPL-3.0 and transforming from a “better terminal” into what it calls an “agentic development environment.” The project hit 37,000+ stars within days and climbed to #2 on GitHub trending this week. Used by 700,000+ developers, Warp now offers the first full-featured open platform for building and orchestrating AI coding agents—with native support for Claude Code, Codex, and Gemini CLI. For developers already using AI agents, this is a major shift: you now have an open source alternative to closed-source IDEs from big tech.

What Makes Warp “Agentic”

Traditional terminals treat their viewport as a single two-dimensional grid of character cells. However, Warp breaks this model with what it calls “blocks”—typed, self-contained units of content that stack vertically. The key innovation: Warp’s BlockList doesn’t care what’s inside a block, only its height. This abstraction allows agent conversation UIs, reasoning steps, and code diffs to coexist seamlessly with traditional terminal output in one continuous scroll stream.

This isn’t just architectural cleverness. Moreover, if you use Claude Code, Codex, or Gemini CLI, Warp gives you a native UI without leaving the terminal: vertical tabs for multi-threaded agent workflows, system notifications so you can context-switch while agents work, and built-in code review tools. Traditional terminals force you to watch agent output scroll by. In contrast, Warp turns that output into structured blocks you can navigate, review, and act on.

Performance matters here too. Specifically, Warp uses a SumTree data structure for O(log n) viewport lookups and virtualizes rendering at two levels—only blocks intersecting the viewport render, and within those blocks, only intersecting rows. The result: agent workflows don’t slow down your terminal, even with long-running sessions.

Tutorial: Setup Warp with Claude Code

Let’s get practical. Here’s how to install Warp and integrate it with Claude Code in under five minutes.

Step 1: Install Warp

On macOS:

brew install --cask warp

Or download directly from warp.dev and drag to Applications.

On Linux (Debian/Ubuntu):

sudo apt install ./warp-terminal.deb

For other Linux distributions, download the AppImage:

./Warp-*.AppImage

Warp supports bash, fish, zsh, and PowerShell.

Step 2: Run Claude Code Inside Warp

Launch Warp and run your standard Claude Code command:

claude code

Warp auto-detects CLI agents and enhances the UI automatically. Consequently, you’ll see agent responses in structured blocks, not raw terminal output.

Step 3: Configure MCP Servers

Warp auto-discovers MCP configurations from ~/.claude.json, .mcp.json, or .codex/config.toml. No reconfiguration needed—Claude Code, Codex, and Warp share the same MCP server set. Furthermore, when you connect an MCP server to Warp, those tools become available to Warp’s Agent Mode alongside shell commands.

Step 4: Enable Notifications

Go to Settings > Notifications > Enable system notifications. Now you can context-switch while Claude Code works in the background. Warp alerts you when the agent needs attention.

That’s it. Five minutes, copy-paste commands, and you’ve upgraded your Claude Code workflow.

Cloud Agents: The Oz Platform

Here’s where Warp differentiates from traditional terminals. Oz is Warp’s orchestration platform for running cloud agents triggered by webhooks, cron jobs, or APIs. This enables autonomous CI/CD workflows without you sitting at a keyboard.

Real-world examples:

Post-deploy monitoring: Set up a webhook on your DeployHQ deploy. An Oz agent watches your error rate for 10 minutes post-deploy and posts to Slack if anything spikes. Six-line prompt plus a webhook URL.

Fix flaky tests: When CI reports a test flake more than twice in a week, an Oz agent triggers automatically, debugs the issue, and proposes a fix via PR.

Dependency hygiene: Schedule a cron job for Friday evenings. An Oz agent opens a Renovate-style PR with patch updates every week.

Sentry auto-triage: A Sentry alert triggers an Oz agent that pulls the stack trace, finds the relevant file, and drafts a PR with a proposed fix.

Deploy an Oz agent with a single command:

oz schedule --skill "flag-cleanup" --cron "0 2 * * *"

Cloud agents run in containerized environments and record everything for audit. Essentially, this is infrastructure automation that doesn’t require Zapier or n8n.

The Open Source Controversy

Warp’s open source move sparked heated debate on Hacker News. Critics pointed out that Warp built on Alacritty, an open source terminal, and raised $50 million without contributing back to the ecosystem. One commenter suggested the company could donate some of that funding if it was genuinely grateful for the foundation it built on.

Others questioned whether “agentic development” is true community collaboration or just a company automating its engineering at scale while routing contributions through proprietary infrastructure.

The community response: OpenWarp, a fork that lets developers plug in any AI provider—DeepSeek, Ollama, OpenAI, Anthropic, local models—with keys staying local. This forced Warp founder Zach Lloyd to respond publicly on Hacker News, saying the official team is “actively exploring bring-your-own-model support.”

This is open source working as intended. Community pressure drives accountability. Warp went AGPL-3.0 to prevent proprietary forks, but it can’t prevent community alternatives that offer more flexibility. Competition keeps vendors honest.

Warp vs Cursor: When to Choose Warp

Warp is terminal-first. Cursor is IDE-first. Both are “agentic,” but they solve different problems.

Choose Warp if:

• You live in the terminal
• You use Claude Code, Codex, or Gemini CLI
• You automate DevOps or CI workflows
• You need cloud agents triggered by webhooks or cron

Choose Cursor if:

• You need deep IDE integration
• You rely on semantic codebase search over SSH
• Your workflow is edit-first, not command-first

Warp’s Oz platform orchestrates parallel cloud agents at the terminal layer. Meanwhile, Cursor’s Composer model powers up to 8 parallel agents within an IDE-first editing experience. The trade-off is operational breadth (Warp) versus deep in-editor integration (Cursor).

For terminal-centric developers, Warp offers something Cursor doesn’t: Full Terminal Use, where agents can attach to PTY sessions, read live terminal buffers, and interact with running processes like database shells, debuggers, and REPLs.

Takeaways

Warp is the first full-featured open source platform for agentic development. If you use Claude Code or other CLI agents, it’s worth trying. The Block Model architecture, MCP auto-discovery, and Oz cloud agents solve real problems—notifications while agents work, structured UI for agent output, and autonomous CI/CD workflows.

The open source move isn’t perfect. Community skepticism about vendor lock-in is healthy, and the OpenWarp fork proves that open licensing enables competition. Nevertheless, for developers who live in the terminal and want native AI agent support, Warp is now a viable alternative to closed-source IDEs.

Next step: Clone the repo at github.com/warpdotdev/warp, install it, and deploy your first Oz cloud agent.