Axonius integrates with the Claude Compliance API to help teams safely adopt Claude across the enterprise

Today, we announced the release of an Anthropic adapter in the Axonius Adapter Network, using the Claude Compliance API to fetch Claude Enterprise artifacts into the Axonius Asset Cloud. The integration lets customers model their Claude footprint against the devices, identities, applications, and infrastructure that make up the rest of their IT environment.

Organizations of all kinds are adopting Claude at breakneck speed across engineering, operations, go-to-market, and more. Security and IT teams know how to onboard a new tool safely, but this one hits a bit differently. Many people were already running Claude on their own. Its footprint touches your SaaS, software, and identity estate in new ways, and AI agents in an enterprise environment are unpredictable by design. Safe AI adoption demands a fresh security model.

That security model must reach past Claude itself. It’s one system in a much larger environment, and the risks that matter rarely exist in isolation. They live in the intersections, such as the person already running Claude on their own, the user who left the company months ago, or the active key on a device with no endpoint protection. Those only resolve when the Claude footprint is set against everything else you run.

That's decision-grade asset intelligence, and the foundation that safe AI adoption rests on.

How the Anthropic adapter works

The Anthropic adapter authenticates to the Claude Compliance API and fetches Claude Enterprise artifacts: users, groups, roles, permissions, and API keys.

Each artifact lands in the Axonius Asset Cloud as a first-class asset, attributed to Claude as its source. The reconciliation layer normalizes it and models its relationships to everything else in your environment, including identity accounts, devices in your endpoint inventory, connected software and SaaS, cloud/on-prem infrastructure, and more. 

Net-net: a Claude user resolves to the person behind it, the devices they use, and the access they have.

The first unlock: discovering shadow AI

You can't protect what you can't see, and security and IT teams are more in the dark than they've ever been. Claude adoption rarely waits for procurement: people sign up on their own, install Claude Code on a work laptop, and start building. Shadow IT isn't new, but shadow AI has made it faster, broader, and higher-stakes, because the tools now reaching into your data and code are the ones nobody approved.

As organizations adopt Claude Enterprise to get a handle on this, the Claude Compliance API tells you the managed users in your tenant. Seeing what it can't report on is the Axonius advantage. We discover Claude and Claude Code usage from the software installed across your devices, then compare it against the managed user population. The negative space between them is your shadow AI: people running Claude on company machines who never appear in the tenant.

That gap is a migration list. Each unmanaged user is someone to bring into the governed tenant, where enterprise controls, data handling, and retention apply. Because reconciliation already ties each person to their device and identity, you know whose machine it is and how to reach them.

The second unlock: reducing the attack surface

Once Claude is reconciled into Axonius' model of every asset and how they connect, you can run it through the same risk assessment as everything else. The findings that matter rarely come from a single fact. One signal on its own is easy to dismiss; risk shows up when several line up at once.

Take one engineer's laptop. Four systems each report something unremarkable on its own. Claude in use, an account outside the managed tenant, a device without EDR, or access to private repositories. Reconciled onto one record, they climb from informational to urgent.

That urgent reading only surfaces because all four land on the same asset record at once. This is the everyday Axonius use case, now extended to your Claude footprint: flagging drift, coverage gaps, and excessive access across the entire environment, with each notable finding tied to the owner who can fix it with all the context to make a decision.

Safe AI adoption starts with the fundamentals

There’s no slowing down AI adoption; that much is clear. Security and IT teams have always had to deal with the fact that every tool you add widens the attack surface: new identities, access, endpoints, and API keys. But pair an expanding attack surface with exploding vulnerability discovery, and it might feel like AI sprawl is too daunting to deal with.

The good news is that the fundamentals are back in fashion, for all the right reasons. For years, the basics like having an asset inventory and closing coverage gaps lost the spotlight to whatever was new and hot. It might sound counterintuitive, but the rise of AI puts the “boring” fundamentals right back at the center where they belong. 

This glorious rise of the fundamentals isn’t unique to the asset intelligence category. Anthropic’s own Zero Trust for AI Agents guide highlights a different problem (i.e., securing autonomous agents), but reaches a parallel conclusion: getting the fundamentals right matters more than having the most advanced AI. 

Say it with us: Safe AI adoption is a fundamentals problem before it’s an agents problem. Whatever AI security strategy you build next starts from the ground truth that you can see and trust.

Getting started

To enable the Anthropic adapter, add it to your Axonius instance and configure the connection for read access to your Claude Enterprise workspace. 

Full setup steps can be found in the Axonius documentation.

Availability

The Anthropic adapter will be generally available to customers through the Axonius Adapter Network on June 15, 2026.

Additional resources

Discover why every AI-era security strategy starts with asset intelligence.

Read the press release and datasheet to learn more about the Anthropic integration.