Claude Apps Gateway: Keeping Control of the AI Your Employees Use

Anthropic has been shipping announcements all summer. After Claude Sonnet 5 and the return of Fable 5 worldwide, the company has now delivered something far less flashy that lands squarely on organizations: the Claude apps gateway, a gateway that lets you take back control over how your teams use Claude. Announced in late June 2026, it targets a very concrete corporate problem, governance.

For a smaller business, the question is no longer whether employees are using artificial intelligence. They already are, often with their own accounts. The question is whether you know who is using it, what it costs every month, and whether you can cut access quickly when someone leaves. That is exactly the hole this gateway fills.

1. What Anthropic just announced

The Claude apps gateway is what is called a control plane: software a company deploys on its own side that sits between its employees and Claude. Technically it is a single lightweight container running on Linux, backed by a PostgreSQL database, and shipped directly inside the tool developers already install. Nothing heavy to stand up, in other words.

According to Anthropic, the gateway solves three irritants that existed until now. Previously, running Claude Code on the cloud platforms meant creating individual access for every person, distributing settings by hand on each machine, and building your own tools to track spending. All of that becomes centralized. The announcement was picked up by several trade outlets, including DevOps.com, which describes it as a fundamental change in how this access is managed.

It works with Claude through Amazon Bedrock, Google Cloud and Microsoft Foundry, as well as through direct access to Claude’s programming interface, with the ability to switch from one provider to another during an outage.

2. The real problem: AI with no guardrails

In plenty of companies, AI adoption came in through the back door. One employee opens an account, another pays with their own card, a third pastes internal documents into a tool nobody at the office ever approved. It moves fast, and it creates an uncomfortable grey zone for an owner.

This is not theoretical. In a survey cited by Anthropic when it launched its offering for small businesses, half the owners questioned named data security as their main hesitation about AI. The brake is not a lack of interest, in other words, it is a lack of governance.

A gateway like this one brings AI back inside the usual corporate security framework. Think of it as what you already do for other software and access: a single entry point, clear rules, a record of what happened. That is the logic we put in place for our clients through our managed IT services.

3. How the gateway works

The heart of the system is single sign-on, or SSO. Instead of everyone keeping an access key on their own machine, employees sign in with the company identity. The gateway plugs into the most common identity directories:

• Google Workspace
• Microsoft Entra ID (the identity behind Microsoft 365)
• Okta
• Any provider that follows the open OIDC standard

Another important point: the gateway hands out short-lived tokens rather than permanent access. In practice, a session lasts about an hour by default. That means a forgotten access does not linger indefinitely, which lowers the risk when a device is lost or stolen. For a business with no full-time security team, that kind of default behaviour makes a real difference.

4. Costs under control, person by person

This is often where AI stings: the invoices climb but nobody really knows who is consuming what. The gateway tags every request with usage measurements and attributes costs by user. From there, the company can set spending caps:

• by day, by week or by month
• at the level of the whole organization, a group or a single person

For a business owner, that turns a fuzzy expense into a predictable budget line. You decide in advance how much the sales or accounting team can consume, and you do not get a surprise at the end of the month. Usage data is also sent to the company’s own infrastructure, which keeps visibility in-house.

5. When an employee leaves: cutting access in one move

The scenario is a classic. An employee leaves, and three months later you realize they still had access to a tool, a shared mailbox or a key somewhere. With the gateway, that risk drops sharply. Because access runs through the identity directory, you simply disable the person there. Their access to Claude expires on its own within the session length, an hour by default.

Anthropic sums the idea up simply: no key cleanup, no orphaned access floating around. It is the kind of basic hygiene everyone knows in theory but that often slips through the cracks in a small team’s daily routine. Mitch Ashley, an analyst at the Futurum group, does note a limit: this gateway controls who has access and at what cost, but it does not yet govern what AI agents do once they are running. Governing the use itself remains a separate project.

6. What this means for your business

You do not need developers or Claude Code to take the lesson from this announcement. The message applies to any company that lets employees use AI tools: governed access beats a patchwork of personal accounts. Three habits are worth starting today:

• Inventory the AI tools actually in use in your business, including the ones nobody officially approved.
• Tie that access to your company identity (Microsoft 365 or Google Workspace) rather than to personal accounts.
• Give yourself a simple way to cut access when someone leaves, without chasing a list of keys.

That is precisely the kind of framework we put in place for businesses in Trois-Rivières and the Mauricie. If you are wondering where to start with AI without opening a security gap, a review of your current access is a good place to begin. Talk it over with our team through the contact page, or see everything covered by our managed IT services.

Frequently asked questions

Is the Claude apps gateway only for large enterprises?

No. Technically it fits in a single lightweight container backed by a database, which puts it within reach of a modest organization that already has some infrastructure. The idea of governing access applies to companies of every size.

Does it replace my Claude subscription?

No. The gateway is a tool for managing access and costs. It sits on top of your use of Claude through the cloud platforms or direct access, it does not supply the AI model itself.

How do I find out which AI tools my employees already use?

An audit of access and software spending usually surfaces the undeclared accounts and subscriptions. It is a simple step a managed IT partner can run quickly, before you decide how to centralize everything.

Govern AI before it governs you

The arrival of the Claude apps gateway confirms an underlying trend: AI is moving from an individual gadget to a corporate tool, with the security rules that come with it. A smaller business does not have to build everything itself to benefit, it just has to apply the right governance habits. To take stock of your access and build a clear framework, take a look at our managed IT services or write to us through the contact page.