What is an agentic development environment (ADE)?

Published July 28, 2026 · by the AQ team

An agentic development environment (ADE) is a software workspace designed around AI coding agents doing the work, with the human directing, reviewing, and stepping in, rather than around a human typing into an editor with AI assistance bolted on. The term was coined by Warp, which launched Warp 2.0 in June 2025 as "the first Agentic Development Environment." Since then the label has spread well beyond one product, and vendors now use it to mean noticeably different things, so any precise definition has to start with the history and then separate the shared core from the marketing.

The shared core of the definition

Strip away each vendor's framing and the definitions converge on one architectural claim: in an ADE, the primary unit of work is a delegated task, not an open file. The human states intent, an agent plans and executes (editing files, running commands, reading output, retrying), and the human reviews the result, usually as a diff or a pull request. An IDE optimizes the inner loop of a person editing code; an ADE optimizes the outer loop of a person running one or more agents.

Most implementations, whatever they call themselves, share a recognizable anatomy:

• A task surface. Some list, board, or manager view of the agent runs in flight, because there is usually more than one.
• Isolation per task. Each agent works in its own checkout, typically a git branch or git worktree, so parallel runs cannot trample each other.
• An agent runtime. Somewhere real for the agent to execute: a shell, a filesystem with the repository, and the tools the project needs.
• Escape hatches. An embedded editor, terminal, and diff view for when the human needs to intervene mid-run rather than wait for the finish.
• A review layer. The run ends in something a human can accept or reject, most often a pull request.

Where the term came from: Warp, June 2025

Credit for the coinage belongs to Warp. The company spent its first years building a fast, modernized terminal, then relaunched in June 2025 as Warp 2.0, positioning itself as neither a terminal nor an IDE but the first Agentic Development Environment. Founder and CEO Zach Lloyd's argument was that agentic work needed a product natively designed for prompting, multi-threading, agent management, and human collaboration with agents across real codebases, and that neither the IDE nor the terminal was that product. Warp 2.0 folded code editing, agent management, terminal use, and shared context into one app.

The bet aged well as a naming exercise. TIME named Warp's ADE one of its Best Inventions of 2025 in the AI category, which pushed the term into mainstream coverage. In April 2026, Warp open-sourced its client and began building it in the open, with cloud agents doing much of the implementation through its Oz orchestration platform, keeping the ADE label at the center of its identity. As of July 2026, warp.dev still leads with the tagline "The Agentic Development Environment."

How the term spread, and where it got contested

Once the category had a name and a TIME award, other vendors moved onto the territory, each bending the definition toward what they had built. As of July 2026:

The contested part is not whether agents matter; everyone agrees on that. It is which layer deserves the name. Warp uses ADE for an all-in-one desktop app. GitKraken uses it for a cross-repo orchestration layer that drives other vendors' agents. Google avoids "environment" and says "platform" for something that spans an editor and an agent manager. Meanwhile "agentic IDE" persists as the label for editor-first products like Cursor and Windsurf, where an agent mode lives inside a conventional IDE. The useful distinction that has emerged: an agentic IDE keeps the editor as the primary surface and adds agency; an ADE makes the agent run the primary surface and keeps an editor as an escape hatch.

What an ADE is not

Three neighboring terms get conflated with ADE, and keeping them apart makes vendor pages much easier to read:

• An agentic IDE is not an ADE. Cursor and Windsurf are editors with strong agent modes. The primary surface is still a file you are editing. The distinction is architectural, not a quality ranking.
• A coding agent is not an ADE. Claude Code, Codex CLI, and similar agents are the workers. An ADE is a place where those workers run, get supervised, and get reviewed. Several ADEs (GitKraken's Kepler, for example) explicitly drive third-party agents rather than shipping their own.
• A cloud agent platform is not automatically an ADE. Fire-and-forget services that take an issue and return a pull request skip the supervision layer. Most ADE definitions include the ability to watch and steer a run in progress.

Questions to ask any product wearing the label

Because the term is young and contested, the label alone tells you little. The concrete questions that separate the implementations:

• Where do agents execute? On your laptop (dies when the lid closes), in the vendor's cloud (custody and compliance questions), or on infrastructure you control.
• What isolates parallel tasks? Worktrees, containers, separate VMs, or nothing.
• Who can see a running session? Most ADEs are single-player: the run belongs to whoever started it, and teammates see a pull request at the end, if anything.
• Whose model account pays? Bundled vendor pricing with a margin on tokens, or your own existing Claude, OpenAI, or other subscriptions.
• What happens to a run when you disconnect? Persistence is the difference between delegating work and babysitting it.

Where AQ fits

AQ is the multiplayer coding harness where engineering teams run AI coding agents like Claude Code and Codex together: shared live terminals, a code editor, and app previews, in your own cloud. In this page's terms, AQ matches the ADE anatomy (workspaces as the task surface, one isolated git worktree per workspace, live terminals and an editor as escape hatches, pull requests tracked per workspace) while answering the questions above differently from the desktop-app ADEs. Agents run as the real CLIs (Claude Code, Codex, Cursor Agent, Kimi, Grok, or plain shells) in persistent tmux sessions on your team's VM, so a run survives a closed laptop and resumes from any device. Sessions are multiplayer: teammates open the same workspace and see and steer the same live session, rather than finding out at the pull request. Each engineer signs into the CLIs with their own accounts, so AQ never marks up model usage, and there is no shared multi-tenant execution tier.

Plainly: if what you want is a single desktop app for one developer supervising agents, Warp defined this category and is a strong default, and Google's Antigravity is the platform-scale take on the same idea. AQ's territory is the version of this category those products mostly skip as of July 2026: agent sessions that are durable, shared across a team, and executing on infrastructure you own. The Free plan is the full product on a VM you bring (you pay only your cloud provider); the Team plan is $100 per user per month (promotional; standard $200, billed monthly) and includes a dedicated always-on VM managed by AQ in its own isolated network, with the rate locked for your first 12 months.