Google Bundles Antigravity Into Enterprise Seats • The Weighted Average

Google has folded a $200-per-month coding agent into subscriptions companies already buy, and the interesting part is the billing plumbing rather than the agent. Antigravity is now available through eligible Gemini Enterprise Standard, Plus, and Standard Emerging Market subscriptions, with administrators enabling it per licensed user from the Gemini Enterprise admin console, according to ITBrief’s report on the change. The package adds monthly project-level budget caps in Google’s billing console, shared token pools so one team’s unused quota serves another, opt-in overage billing with spend caps, and central tracking of token consumption, API calls, and developer activity.

Price the delta. Antigravity’s own pricing page lists a free Individual tier and paid access through Google AI Pro and Ultra plans, with an Organization plan billed by consumption under Google Cloud terms and “included in select Gemini Enterprise app subscriptions.” Independent tracking of the tiers puts the top Ultra plan at $200 per month, cut from $250 in a May 2026 restructuring, per Emergent’s breakdown of Antigravity pricing. That is $2,400 per developer per year in list terms — the number a platform lead no longer has to defend as a separate purchase order once the capability rides an existing seat. Gemini Enterprise Standard and Plus seats themselves start around $30 per user per month on Google’s published edition comparison, which is where the bundling math gets uncomfortable for standalone coding-agent vendors.

Governance was the actual blocker

The Register’s account is explicit that this shipped because customers asked for it: enterprises pushed Google to add security, compliance, and governance controls and to put Antigravity inside the tools developers already use, per The Register’s report on the enterprise controls. The controls are the concrete part. File access can be configured to deny everything outside the agent’s working folders. Terminal commands can be set to always ask, always sandbox, or always proceed. Browser access can be granted, denied, or restricted to a website allowlist.

Those three switches are the difference between a pilot and a rollout. An agent that can run arbitrary shell commands against a developer’s machine is a security review; an agent whose command policy is set centrally and logged is a configuration item. Distribution follows: Antigravity now reaches Visual Studio Code, Visual Studio in preview, JetBrains in preview, and Zed in preview, alongside the Antigravity 2.0 desktop app and CLI — four IDE surfaces beyond its own, which removes the last practical objection from teams unwilling to change editors.

There is a subtler distribution effect. An agent installed as an extension inherits the editor’s trust boundary, its existing project configuration, and its telemetry, which means adoption no longer requires a developer to relocate their working environment. Historically that relocation cost was the single largest source of pilot attrition in agentic tooling: teams evaluated a new IDE, not a new agent, and abandoned both.

What this does to everyone else’s price list

The competitive read is that agent capability is becoming a bundled feature of the cloud contract you already signed, the same pattern that made bundled observability and bundled CI hard businesses to stand alone in. A specialist agent vendor now has to clear a higher bar: it must be better than a tool that costs nothing incremental and inherits the customer’s existing identity, billing, and audit stack. That pressure is already visible in how quickly rivals are pushing into collaboration and hosting surfaces rather than competing on the agent loop itself.

The counterpoint is real. Pooled tokens and budget caps are cost controls, not cost reductions — consumption still bills against the pool, and Google’s own materials describe per-user and per-team spending controls as arriving later. Until those land, a single runaway agent session can drain a project pool that other teams are relying on, which is a worse failure mode than an individual seat hitting a personal limit. Teams that have built discipline around per-developer AI spend controls will find the pooled model a step backwards in attribution even as it improves procurement.

The operator call is narrow. If you already hold Gemini Enterprise Standard or Plus seats and your objection to agentic coding was procurement or policy rather than quality, enable it this quarter with terminal commands set to sandbox and browser access restricted, then measure token burn per merged pull request before expanding. If you do not hold those seats, this changes nothing about the agent’s quality and everything about the price you should be quoting rivals. And if your organization is building its own vertical capability rather than buying, note the direction of travel in today’s lead on Harvey’s in-house post-training run: the platform layer is consolidating downward into the cloud bill, which raises the value of anything the bundle cannot supply.

What would change the verdict: published per-seat consumption data. Bundling wins if the effective cost per merged change beats a standalone subscription. If pooled overages routinely exceed what the same team spent on dedicated seats, the bundle is a discovery mechanism for a larger bill, and the control-plane discipline the pull request already provides matters more than the license it runs under.