Does Google AI have a vscode extension?
Yes. Google's VS Code extension is Gemini Code Assist. As of August 2, 2026, the extension remains available to Standard and Enterprise license holders: install it, sign in to Google Cloud with the licensed account, and select an enabled project. Google ended IDE-extension and Gemini CLI requests for Gemini Code Assist for individuals, Google AI Pro and Google AI Ultra on June 18, 2026, removing their consumer "Login with Google" route.
Why — the first-principles explanation
The durable answer is yes, the VS Code extension exists. Search VS Code Extensions for Gemini Code Assist. It provides code completion, code generation, chat and an agent mode inside VS Code. The product did not disappear when Google changed consumer access.
What changed is who can authenticate through which route. Google's dedicated consumer-account deprecation page says that, starting June 18, 2026, Gemini Code Assist IDE extensions stopped serving requests for Gemini Code Assist for individuals, Google AI Pro and Google AI Ultra. The same change applies to Gemini CLI, and the consumer Login with Google option is no longer available. Google directs those users to the Antigravity family of products. The page explicitly says Standard and Enterprise access remains unchanged. That official notice supports the cutoff; it should not be softened into an unsupported rumor or exaggerated into "Google no longer has a VS Code extension."
For an organization, the current path is Gemini Code Assist Standard or Enterprise. An administrator purchases the edition, assigns or enables license assignment, enables the Gemini for Google Cloud API in a project, and grants the required IAM access. A developer then installs the extension, clicks Login to Google Cloud, signs in with the licensed account and selects the configured Gemini Code project. This Google Cloud sign-in flow is distinct from the removed consumer Login with Google route.
Standard and Enterprise share the main IDE experience. Standard is the normal organizational coding-assistance tier with enterprise data controls. Enterprise includes Standard's capabilities and adds private-repository code customization plus broader Google Cloud integrations. A Google AI Pro or Ultra subscription is not a substitute for either organizational license under the current deprecation notice.
Agent mode is a layer inside the extension, not a separate VS Code extension. It can inspect workspace context, read and write files, run terminal commands, use Google Search or fetched web content, and connect to MCP servers. It normally asks before mutating files or resources unless permissions were broadened; read-only tools might run without a prompt. Google documents agent mode as powered by Gemini CLI and exposing only a subset of the CLI experience in VS Code. Gemini CLI itself is a separate open-source terminal agent. For Standard and Enterprise, its privacy protections apply and its model-request quota is shared with VS Code agent mode.
The quota numbers need labels. Google currently documents a combined agent-mode-and-CLI maximum of 1,500 model requests per user per day for Standard and 2,000 for Enterprise. One prompt can trigger multiple model requests, so those figures are not the same as prompts per day, and availability can still be constrained at high demand. The general Gemini Code Assist quota page also lists 2 requests per second per user in a project, 6,000 code requests per day and 960 chat or other Cloud Assist responses per day. If agent quota is exhausted, Google documents an optional Gemini API or Vertex AI API key in VS Code; that is a separate API-usage path, not restoration of the retired consumer login.
Treat the extension as a capable code-changing client, not a harmless autocomplete. Standard and Enterprise prompts and responses are not used to train Google's underlying models, and Google describes the service as stateless unless a customer configures logging. But the service can receive prompt text, conversation history, open or adjacent file snippets and cursor context, while product telemetry is also collected. Keep secrets out of prompts and reachable files, restrict tools and MCP credentials, leave automatic approval off unless the workspace and command set are deliberately trusted, review every diff, and run tests, linters and security checks before merging generated code.
An example that makes it click
Two developers install the same Gemini Code Assist extension. Mei works at a company with a Standard license. Her administrator has enabled the Gemini for Google Cloud API and assigned the required access, so she clicks Login to Google Cloud, selects the approved project and sees completions, chat and the Agent toggle. Alex has a personal Google AI Pro subscription. The extension installs, but the old consumer login no longer grants service; the official deprecation page sends that account type to Antigravity instead. The installation is identical, but entitlement and authentication determine whether it works.
How to do it
- Check entitlement before installing. Continue with this VS Code flow only if your organization has assigned or can automatically assign a Gemini Code Assist Standard or Enterprise license; individual, Google AI Pro and Google AI Ultra consumer access was retired on June 18, 2026.
- Have an administrator confirm the setup: the subscription and license are active, the Gemini for Google Cloud API is enabled in a Google Cloud project, and you have the Gemini for Google Cloud User and Service Usage Consumer roles or equivalent permissions.
- In VS Code, open Extensions with Ctrl/Cmd+Shift+X, search for Gemini Code Assist, follow the listing linked by Google's setup documentation, click Install and restart VS Code if prompted.
- Click Gemini Code Assist in the activity bar, choose Login to Google Cloud, allow the browser sign-in and authenticate with the same Google account that holds the organizational license.
- Use the Gemini Code Assist status menu to select the Gemini Code project whose API and billing setup your administrator approved. If the API is not enabled, use the IDE prompt only if you are authorized to enable it.
- Verify the connection: confirm the Gemini icon and selected-project status, open a supported-language file, request a small explanation in chat and check that a completion or response arrives. A license error usually means the account, license, project, API or IAM setup is wrong—not that the extension was discontinued.
- Test agent mode on a disposable branch with a bounded task. Inspect its proposed plan, deny unnecessary tools, review every file and command, and run the repository's normal tests before accepting the result.
- Track the correct limits. Agent mode and Gemini CLI share the edition's daily model-request pool; one prompt can consume multiple requests. Check the current official quota page rather than treating 1,500 or 2,000 as a permanent prompt count.
- Before production use, document allowed repositories, data classification, logging, MCP servers, credential handling, approval rules, test gates and rollback ownership. Recheck Google's setup, deprecation, quota and privacy pages when renewing or changing editions.
Key facts
- Gemini Code Assist is Google's supported VS Code extension for code completion, generation, chat and agentic assistance.
- Starting June 18, 2026, Google stopped serving IDE-extension and Gemini CLI requests for Gemini Code Assist for individuals, Google AI Pro and Google AI Ultra and removed the consumer Login with Google path.
- The same official deprecation notice says Gemini Code Assist Standard and Enterprise access to the IDE extensions and Gemini CLI remains unchanged.
- Standard and Enterprise users authenticate to Google Cloud with a licensed account and select a project with the Gemini for Google Cloud API enabled; licensing, API enablement and IAM are separate prerequisites.
- Enterprise includes Standard's IDE capabilities and adds private-code customization and broader Google Cloud integrations.
- VS Code agent mode is powered by Gemini CLI and exposes a subset of CLI functionality; Gemini CLI remains a separate terminal agent.
- Agent mode can use files, terminal commands, search, web content and MCP servers. Mutating actions normally require permission, while read-only actions might not prompt unless tools are restricted.
- Google documents combined agent-mode-and-CLI daily maximums of 1,500 model requests per Standard user and 2,000 per Enterprise user; one prompt may cause multiple model requests.
- Google's general quota page also lists 2 requests per second per user in a project, 6,000 code requests per day and 960 chat or other Cloud Assist responses per day.
- Google says Standard and Enterprise prompts and generated responses are not used to train or fine-tune its underlying models; customer-configured logging, file context and service telemetry still need governance.
- Agent mode does not provide the standard chat's recitation/source-citation behavior, so users cannot rely on an in-agent citation control as a substitute for code review and license checks.
- An optional API key after agent-quota exhaustion is a separate Gemini API or Vertex AI usage path and does not reinstate the retired consumer Login with Google entitlement.
Choose the coding workflow, not just the extension
Confirm entitlement, data controls, tool permissions, shared quotas and accepted-change cost, then compare Gemini Code Assist with the same repository tasks and tests used for other coding tools.
▶ The 60-second explainer (script)
Does Google AI have a VS Code extension? Yes. It is called Gemini Code Assist, and the extension still exists. What changed is access. Google's current deprecation notice says that on June eighteenth, 2026, the IDE extension and Gemini CLI stopped serving Gemini Code Assist for individuals, Google AI Pro and Google AI Ultra. Their consumer Login with Google route was removed, and Google points those users to Antigravity. Standard and Enterprise licenses are explicitly unaffected. For those organizational tiers, install Gemini Code Assist from VS Code Extensions, click Login to Google Cloud, use the licensed account and select a Google Cloud project with the Gemini for Google Cloud API enabled. Standard covers the core organizational coding workflow; Enterprise adds private-repository code customization and broader Cloud integrations. Inside VS Code, agent mode can read files, write code, run commands and use MCP servers. It is powered by Gemini CLI but exposes only part of the separate terminal agent. Agent mode and Gemini CLI share a daily model-request allowance: Google currently lists fifteen hundred for Standard and two thousand for Enterprise, and one prompt may consume several requests. Keep automatic approval off by default, restrict tool access, protect secrets, inspect every diff and run your normal tests. So the precise answer is: yes, Google has a VS Code extension; consumers lost the old login path, while licensed Standard and Enterprise teams still use it.
What authoritative sources say
People also ask
What is Google's VS Code AI extension called?
Gemini Code Assist. Google's current documentation supports it in VS Code and describes code completion, generation, chat and agent mode.
Is Gemini Code Assist still available in VS Code?
Yes. The extension remains available. Current service access is for Gemini Code Assist Standard and Enterprise licenses; the consumer editions lost their IDE service and Login with Google path on June 18, 2026.
Is Gemini Code Assist free for individuals?
Not through the old VS Code consumer path. Google's deprecation notice says Gemini Code Assist for individuals stopped serving IDE-extension and Gemini CLI requests on June 18, 2026. Check Antigravity for Google's current consumer coding route.
Does Google AI Pro or Google AI Ultra unlock Gemini Code Assist in VS Code?
No under the current notice. Google lists both AI Pro and AI Ultra among the consumer tiers whose IDE-extension and Gemini CLI requests stopped on June 18, 2026. A Standard or Enterprise organizational license is a different entitlement.
Why does Gemini Code Assist say I do not have a valid license?
First confirm that you signed in with the licensed organizational account. Then verify the Standard or Enterprise license assignment, selected Google Cloud project, Gemini for Google Cloud API and required IAM roles. Installing the extension alone does not create an entitlement.
What is the difference between Standard and Enterprise?
Both provide the core supported IDE assistance and enterprise controls. Enterprise adds private-source-code customization and broader Google Cloud integrations. Evaluate those needs before paying for the higher edition.
Is Gemini Code Assist agent mode the same as Gemini CLI?
No. Gemini CLI is a separate terminal agent. Google's VS Code agent mode is powered by Gemini CLI but exposes a subset of its functionality inside Gemini Code Assist, and the two share the edition's agentic request quota.
What are the Gemini Code Assist agent and CLI limits?
Google currently lists a combined maximum of 1,500 model requests per Standard user per day and 2,000 per Enterprise user per day. One prompt may generate multiple requests, and these values can change, so use the official quota page for purchasing or capacity planning.
Can an API key restore the old consumer login?
Google documents an optional Gemini API or Vertex AI API key for continuing agent-mode use after its daily quota is exhausted. That is a separate API-usage path, not a reinstatement of the retired consumer Login with Google entitlement.
Does Google train on my company's code and prompts?
Google says Standard and Enterprise prompts and responses are not used to train or fine-tune its underlying models. The service can still receive file context and collect service telemetry, and customer-configured logging changes retention, so review the full data-governance controls before approving sensitive repositories.
Is agent mode safe to let run automatically?
It can read and write files, run terminal commands and call MCP tools, so treat it like a code-changing operator. Keep least-privilege tool settings, avoid automatic approval by default, review commands and diffs, protect credentials and require normal tests and security checks.