TrendAI, Trend Micro’s AI security unit, has expanded its integration with Claude’s Compliance API so its Vision One platform also captures Claude Code and Cowork sessions from Claude Enterprise customers. With the change, announced on September 29, 2026, security teams can review what each employee asked AI agents, what they answered, and which tools they used, and cross-reference that with the rest of the company’s security signals.
Key facts about the TrendAI integration with Claude Code and Cowork in 20 seconds
- Vision One receives prompts, responses, and tool calls from Claude Code and Cowork.
- Each session is tied to a verified user.
- It only works with Claude Enterprise.
- Claude Code cloud sessions and those run on Bedrock, Vertex AI, or Foundry are excluded.
The expansion builds on the integration TrendAI announced with Anthropic in June 2026, which until now focused on Claude conversations. The problem it’s trying to solve is an increasingly common one: AI agents write code, call tools, and access business systems on behalf of employees. Security teams usually know those agents are being used, but not what was asked of them, what they did, or whether they followed internal policy.
What the security team can see
The Compliance API is the interface Anthropic offers Claude Enterprise customers to pull usage content for auditing and compliance purposes. Since August it also covers Cowork (desktop, web, and mobile) and Claude Code (terminal, desktop app, and IDE). According to Anthropic’s documentation, each transcript includes the user’s requests, the assistant’s responses, tool calls and their text results, along with metadata such as the user’s ID and email, the organization, and timestamps.
TrendAI feeds those records into Vision One, its detection and response platform, where they can be correlated with workstation, identity, cloud, and network events. According to the company, teams will be able to:
- Detect tool calls that fall outside expected limits.
- Link the activity of an AI session to a specific user during an investigation.
- Check whether agent workflows follow required processes.
- Create their own detections around agent behavior.
- Keep an auditable record of AI activity for compliance reviews.
Vision One continues to receive Claude Enterprise activity logs as well, including sign-ins, admin actions, and configuration changes, plus Claude Platform logs covering admin, system, and resource events. The feature is already available to Vision One customers using Claude Enterprise.
Rachel Jin, Chief Platform and Business Officer and head of TrendAI, argues that security teams need transparency to validate agent activity against company policy, and that coding agents and knowledge-work agents are already doing real work inside the business. In her view, they need to be investigable just like any other activity in order to maintain control as these tools spread.
What’s left out
TrendAI’s announcement doesn’t spell out the limits, but the Compliance API documentation does, and they’re worth keeping in mind. Excluded are Claude Code sessions authenticated with a console API key, sessions run through Amazon Bedrock, Google Vertex AI, or Microsoft Foundry, Claude Code cloud sessions, and sessions from organizations with zero data retention. Internal reasoning blocks, system prompts, and images or PDFs are not included either, and tool inputs and outputs are truncated to 10,000 bytes by default.
Another important nuance: for local sessions, the log shows what was sent to and returned from the Claude API, not everything happening on the user’s machine. To see what’s actually happening on the laptop, workstation security tools are still needed, which is exactly what Vision One proposes to correlate against. According to Trend Micro’s documentation, the connection runs through a connector the customer deploys on AWS, which analyzes the data with AI Guard. Transcripts are retained for six years by default, unless the organization sets a shorter period.
Monitoring agents, a growing market
TrendAI isn’t alone in this space. AI agent security has become one of the industry’s most active fronts, as shown by recent moves such as Palo Alto Networks’ acquisition of Koi or Zscaler’s move with Symmetry Systems. For anyone unclear on what each Anthropic product does, we’ve previously explained the differences between Claude Chat, Cowork, and Code.
There’s also a question European companies shouldn’t overlook. These logs tie every request to an identified employee and store the full content of their work conversations with AI. Before turning this on, it’s worth reviewing how that oversight fits with data protection regulations and internal policy, and informing staff about the scope of the monitoring.
Frequently asked questions
What is Claude’s Compliance API?
It’s Anthropic’s interface that lets Claude Enterprise customers pull usage content, including Claude Code and Cowork sessions, for auditing and compliance purposes.
What data does TrendAI Vision One collect from Claude Code and Cowork?
Prompts, responses, and tool calls from each session, tied to a verified user, along with Claude Enterprise and Claude Platform activity logs.
Does it work with any Claude plan?
No. It’s available to TrendAI Vision One customers who use Claude Enterprise.
Which sessions aren’t covered?
According to Anthropic’s documentation, Claude Code sessions authenticated with a console API key, those run on Bedrock, Vertex AI, or Foundry, Claude Code cloud sessions, and sessions from organizations with zero data retention.
