Google is bringing its Antigravity AI coding agent into developers' existing code editors, expanding beyond the standalone desktop application it launched in November 2025. The company announced Thursday that extensions are now available for Visual Studio Code, Visual Studio, JetBrains IDEs, and Zed, while also making the tool accessible through qualifying Gemini Enterprise subscriptions. The move lets programmers hand multi-step engineering tasks to an AI agent without leaving their preferred development environment.
The extensions allow developers to launch agent conversations in a side panel, examine inline code changes, review plans, and assign complex tasks without switching to the Antigravity 2.0 desktop app. The VS Code extension works now on macOS, Linux, and Windows through Microsoft's extension marketplace, while the Visual Studio 2026 extension for .NET projects remains in preview. Google supports JetBrains tools—including IntelliJ IDEA, PyCharm, WebStorm, GoLand, CLion, and Rider—starting with version 2026.2.1, alongside Zed. A single Antigravity account functions across all environments, eliminating separate sign-ins or license management for each editor.
Administrators can activate Google's developer tools for staff on Gemini Enterprise Standard, Plus, or Standard Emerging Market plans, and set monthly spending caps for each Google Cloud project. When the included allocation runs dry, Antigravity either stops working or shifts to pay-as-you-go pricing. Controls for individual users and teams are scheduled to arrive later this year. Until then, everyone in the same edition, project, and region pulls from one shared credit pool. Google describes the allowance as monthly but tracks it using a rolling seven-day pool, and any unused credits vanish when the pool resets. A nontrivial engineering job can burn through 150,000 to 200,000 tokens, while multi-agent handoffs pile on additional input tokens each time work moves between agents. One built-in Claude Code skill was recently discovered loading more than 200,000 tokens before responding to a question. Without a default allocation per developer, an agent-heavy workflow could exhaust the team's entire pool in hours. Microsoft recently rolled out AI token budgets for its engineering divisions after learning that many of its engineers were spending hundreds to thousands of dollars monthly on tokens. Uber reportedly drained its entire 2026 AI coding budget in the first four months of the year, and an internal Amazon project matching author records with product listings exceeded its planned budget by 860%.
Enterprise developers sign in with company credentials, then select the Google Cloud project and region that Antigravity should use. Organizations can link to their existing identity provider through Workforce Identity Federation, and developers can also authenticate using Application Default Credentials. Once authenticated, agent sessions adopt the organization's IAM policies, VPC Service Controls, and regional data boundaries. Google says data from enterprise Antigravity sessions isn't used to train its foundation models. Admins can restrict an agent's workspace, block browser access, and choose which MCP servers it can use—once one of those connections reaches a production system, a misconfiguration can have serious consequences. ElevenLabs' MCP server, for example, let Claude delete production voice agents from a chat window, a reminder that a connection with that kind of authority needs to be managed like any other privileged access.
Placing Antigravity inside existing editors lets Google deploy its agents onto developers' machines without requiring a company-wide tooling overhaul, while Gemini Enterprise keeps every session bound to the same policies and project budget whether it starts in VS Code or JetBrains. Developers can swap editors without altering what the agent can reach or how its usage is billed. The extension gets Antigravity through the door; the control plane dictates what it can do once it's inside. For enterprises weighing the trade-off between developer velocity and operational oversight, the real test will be whether centralized guardrails can keep pace with decentralized adoption. Token budgets and identity controls only matter if teams actually understand their exposure before the bill arrives.