Closing the AI Visibility Gap: Akto's Integration with Claude's Compliance API

Enterprise teams are putting Claude to work at a pace few tools have matched. Developers build with it, analysts draft with it, and increasingly, AI agents act on its behalf through connected MCPs and Skills. As that adoption scales, a familiar problem resurfaces: security and compliance teams have far less visibility into what happens inside Claude than they have for nearly every other application in their stack.

That gap is what Akto's new integration with Claude's Compliance API is built to close.

The blind spot inside everyday AI workflows

Traditional data security tooling watches data at rest and in motion across storage, SaaS, and the network. It was never designed to see what happens inside an AI conversation. Yet that is exactly where sensitive information now flows. An engineer pastes a configuration snippet to debug it. An analyst uploads an internal forecast for a summary. A new MCP server gets wired into a workspace, and a Skill is created that changes what an agent can do.

None of these actions are necessarily malicious. But each one can move regulated data, credentials, or intellectual property through Claude with no alert, no audit trail, and no straightforward way to investigate afterward. As organizations scale agentic AI workflows, where both people and autonomous agents are acting, the volume of activity quickly outpaces any team's ability to review interactions one at a time.

What the integration does

Akto Atlas is an agentic AI security platform for employee endpoints. It integrates with the Compliance API to pull your organization's Claude activity directly into the Akto dashboard, without deploying any endpoint agents. Once connected, Akto continuously ingests activity and surfaces it for discovery, alerting, and reporting, so teams can see who is using Claude, what is being shared, which projects and Skills are active, and which events deserve attention.

Because Akto is purpose-built for agentic AI, the integration pays particular attention to the surfaces where autonomous behavior shows up. Using the Compliance API's Activity Feed and Objects APIs, Akto ingests two kinds of data.

Activity logs

• Chat activity - conversations created, viewed, updated, or deleted

• User and organization management - invitations accepted or rejected, role changes, and group membership

• MCP server events - MCP server additions, removals, and tool policy updates

• Skills activity - Skills created, replaced, or deleted

Conversation content and Projects

• The contents of chat messages, uploaded files, and project content, so sensitive data shared in a conversation can be inspected rather than only counted.

• Projects - Akto discovers Claude Enterprise Projects and their metadata, including ownership, privacy settings, attachments, and chat activity, so teams can see how Projects are being used across the organization and who is responsible for each one.

How it works

Akto polls the Compliance API at regular intervals using your Compliance Access Key, streams the events to its ingestion service, and presents them in the Akto Atlas dashboard. A dedicated Claude Compliance asset appears under Agentic AI Discovery, representing all discovered Claude usage across your organization and connecting it to the rest of your AI inventory.

From there, the integration comes together as three capabilities in the dashboard:

• Audit every AI interaction - Akto ingests chat activity, Projects, MCP server events, and Skills, giving you a complete picture of how Claude is used across your organization

• Flag guardrail violations - Akto inspects the contents of chats, files, and Projects, flagging data leakage (PII, credentials, intellectual property) and other guardrail violations as activity is ingested.

• End-to-end Anthropic governance - Akto flags and blocks risky Skills and MCP servers, surfaces RBAC changes like privilege escalation, and blocks personal-account access through Akto's endpoint and browser controls.

Through the Compliance API, this is detective monitoring: Akto reads activity and raises alerts, and the integration itself does not modify Claude. Enforcement actions - blocking risky Skills, MCP servers, or personal-account access, run separately, through Akto's endpoint and browser controls. From one platform, security, compliance, and IT teams can investigate incidents, report on usage, and apply their organizational policies, so governance scales alongside adoption rather than lagging behind it.

What teams get

Once Claude activity lives in Akto, the questions that used to have no answer suddenly do:

• Security teams can see when sensitive data moves through a conversation, who was involved, and how that activity connects to the rest of the environment - and get alerted on anomalous access before it becomes an incident.

• Compliance teams can produce audit-ready evidence of how Claude is being used, who holds access, and which policies applied, without launching a manual investigation each time.

Getting started

Setting up the integration takes a few minutes:

1. In Claude Enterprise, the Primary Owner enables the Compliance API under Organization Settings → Data and Privacy, then creates a Compliance Access Key with the scopes you need (read:compliance_activities for the activity feed, read:compliance_user_data for chat, file, and project content, and read:compliance_org_data for org metadata, roles, and groups). Scopes are immutable once a key is created, so select what you need up front.

2. In Akto Atlas, open Connectors, find the Anthropic card, and click Connect.

3. Provide your Compliance Access Key, leave the API base URL as https://api.anthropic.com unless you've been given a custom endpoint, add your Akto Data Ingestion Service URL, and click Import.

Akto verifies the key and begins pulling activity. Note that Claude Platform Admin keys provide activity logs only; conversation content - chats, files, and project content- requires a Claude Enterprise Compliance Access Key.

Govern AI as fast as you adopt it

The value of Claude comes from letting teams move quickly. The job of security is to make sure that speed doesn't come at the cost of visibility. By bringing Claude activity, projects, MCP servers, and Skills into one place, and continuously checking them against your policies, Akto's integration with the Compliance API lets organizations extend Claude with confidence instead of limiting adoption out of uncertainty.

To set up the integration or see it in action, reach out to the Akto team or visit the Akto Atlas documentation.