This article was originally published on aicoderscope.com
On April 28, 2026, Warp open-sourced its terminal client under AGPL-3.0, picked up 60,000 GitHub stars, and declared itself an "agentic development environment." OpenAI signed on as founding sponsor. The announcement looked like a triumph of developer-first idealism.
Read the fine print and a different picture emerges: the terminal is free; the product that matters — Oz, Warp's cloud agent orchestration platform — remains fully proprietary. Warp is not becoming an open-source project. It is becoming an enterprise SaaS company with an open-source frontend.
None of that is inherently bad. But it is what this review is actually about: does the $20/month Build plan deliver enough AI value to justify adding Warp to a stack that probably already includes Cursor or Claude Code?
What Warp is in May 2026
Warp's product now has three layers:
Warp Terminal — the terminal client, open-source AGPL-3.0. Rust-based, GPU-accelerated, available on Mac, Linux, and Windows. The core terminal features (blocks, Warp Drive, session sharing, settings file) are free and remain free.
Warp Agent — an AI coding agent embedded in the terminal. Runs locally for interactive work. Handles natural language command generation, code review, debugging assistance, codebase Q&A, and voice input. Consumes credits from your plan.
Oz — Warp's proprietary cloud orchestration platform. Runs agents in the background, coordinates multi-agent workflows, triggers on events from Slack, Linear, or GitHub Actions, and orchestrates third-party CLI agents including Claude Code and Codex. Oz is where the enterprise pitch lives.
Around 1 million developers use Warp as their primary terminal. The pivot to agentic tooling is a bet that those developers will pay to automate their workflows beyond what a local agent session can handle.
Pricing breakdown
Warp simplified its pricing in December 2025, replacing the old Pro/Turbo/Lightspeed tiers with two paid plans.
| Plan | Price | Monthly Credits | BYOK | Repo Indexing | Notes |
|---|---|---|---|---|---|
| Free | $0 | 75 (150 first 2 months) | No | No | All terminal features included |
| Build | $20/mo | 1,500 | Yes | 40 repos, 100k files each | Unlimited Warp Drive, collaboration |
| Business | $50/user/mo | 1,500/seat | Yes | Same as Build | SSO, Zero Data Retention, shared Reload Credits, 50-seat self-serve cap |
Credits power every AI action — command generation, agent turns, codebase indexing queries. Warp does not publish a per-action credit cost publicly, so a 1,500-credit monthly allowance is hard to benchmark without real-world usage. The old Pro plan's overages were priced "eight times more expensive" than included credits; the new Reload Credits ($10–$100 increments, 12-month rollover) are positioned as 50% cheaper than those legacy rates.
BYOK (Bring Your Own Key) on the Build plan is the most interesting option for developers already paying for Anthropic or OpenAI API access. When you use BYOK, queries route directly to your chosen provider — not through Warp's servers — meaning your Warp credits are not consumed. You pay Warp $20/month for the terminal plus orchestration features, and you pay Anthropic or OpenAI separately for the tokens. For a solo developer who already has an Anthropic Max subscription, BYOK turns the $20/month Build plan into a terminal upgrade rather than a competing AI subscription.
The 75-credit free tier is enough to try AI features and evaluate whether the terminal itself is worth using. It is not enough for a daily AI coding workflow.
AI features: what you actually get
Natural language command generation is Warp's most-used AI feature. Type a description of what you want to accomplish ("find all files modified in the last 7 days containing TODO comments") and Warp generates a shell command with an explanation. The generated commands are accurate enough for standard operations; complex pipelines occasionally need editing, same as any LLM output.
Error explanations appear inline when a command fails. Rather than copying an error into a separate chat window, Warp's agent reads the full terminal context and explains the failure with a suggested fix. For developers who spend significant time debugging environment issues, this is genuinely faster than switching to a browser or separate AI tab.
Codebase indexing (Build plan, up to 40 repositories, 100,000 files each) lets Warp Agent answer questions grounded in your actual code. "What does the UserAuth module expect as input?" works without copying the file contents into a prompt. The 40-repo limit covers most solo and small-team setups; enterprise teams with monorepos or hundreds of services will need the Business plan or a conversation with sales.
Voice input lets you speak commands to the agent. Useful in specific workflows (hands busy, voice notes during a terminal session); most developers will not use it daily.
Model selection covers OpenAI, Anthropic, and Google, plus Kimi, MiniMax, and Qwen added at the open-source launch. An "auto (open)" routing option picks the most cost-effective model for a given task. Claude Sonnet 4.6 is the default for coding tasks when using managed credits.
Warp Agent + Oz for automation is the feature class that separates Warp from adding an AI plugin to your existing terminal. Local agent sessions handle interactive work. Oz cloud agents handle background execution: run a nightly code review, respond to a Linear ticket comment, trigger a test suite when a PR is created. These agents can orchestrate Claude Code or Codex sessions, not just Warp's own models. For a team doing agentic workflows at scale, Oz is the actual product.
The terminal itself
Before evaluating the AI, it is worth stating the baseline: Warp is a genuinely good terminal. The Warp Blocks model — grouping each command with its output into a discrete, selectable block — solves the "scroll to find where the last command started" problem that plagues long sessions. Copy just a command's output, not the surrounding noise.
Warp Drive syncs commands, notebooks, and settings across devices. The settings file (programmatic control, cross-device portability) makes onboarding a new machine a one-file operation. Session sharing works for pair debugging without screen sharing.
On performance: Warp's Rust + GPU rendering is fast. Switching from a traditional terminal, the visual responsiveness is noticeable. This is not a marginal improvement.
The AGPL license means the terminal code is auditable and self-hostable, but any modifications to the client distributed to users must also be released under AGPL. This limits commercial forks. Some developers in the open-source community noted that Warp's contribution workflow routes through Oz's proprietary agent infrastructure — a model where the community supplies ideas and validation while Warp's cloud agents do much of the implementation.
How it compares
The honest comparison depends on what problem you are solving.
Warp vs Cursor — these are not substitutes. Cursor is an IDE with a terminal panel. Warp is a terminal with AI features. If you write code in an editor, Cursor's AI (inline completion, Composer, Agent mode) operates where you spend 80% of your time. Warp's AI operates in the 20% that's terminal work. The $20/month Build plan competes with Cursor Pro on price but not on code-generation quality for in-editor work. Most developers who pay for Cursor will find Warp's coding AI redundant for anything above shell scripting.
Warp vs Claude Code — this is more interesting. Both are terminal-based agents. Claude Code runs in any terminal at $20/month (Pro) for agentic tasks. Warp Agent runs in Warp's terminal with tighter environmental context (your actual shell state, codebase index) at the same price. For pure code quality, Claude Code's Opus 4.7 (87.6% SWE-bench Verified) is the stronger model. Warp's advantage is the terminal-native int
Top comments (0)