Your business is running more AI than your security team can see. Agents act on their own and models sit behind every new feature while MCP servers connect them to the data that matters most.
When something goes wrong, most teams can’t answer three basic questions. Who or what acted? What data did it touch? What did it cost? That is the accountability gap that Atlas AIDR closes.
What Is Atlas AIDR?
Atlas AIDR is a new capability inside Atlas, available now to eSentire customers and partners. It turns every instrumented AI session into an attributable record. Each record is tied to a user, a team and a workstation.
There are no additional consoles to staff and no new vendor relationship to manage. Atlas AIDR works with the AI providers and security tools you have already chosen.
Atlas AIDR comes from our acquisition of a stealth-mode AI security startup, and we shipped its technology inside Atlas in weeks. “Every enterprise is racing to put AI to work, and almost none of them can tell you what their AI agents are doing with their proprietary data, or what they cost,” said James C. Foster, CEO of eSentire. “That is the difference between a platform and a portfolio: when we buy something, our customers get it.”
What It Does
AI adoption moved faster than governance. 76% of employees now use AI at work, up from 30% two years ago. 43% of breached organizations had shadow AI involved. The risk is highest in coding agents and desktop applications, so that is where Atlas AIDR goes deepest. These tools carry real credentials and can act without waiting for a human.
Know your AI estate
Atlas AIDR builds a live inventory of the agents and models in use, plus the providers and MCP servers they connect to. Every entry traces back to the people and workstations behind the session. The full session timeline records each prompt and response along with the tool calls in between. For most organizations, the first session is also the first time they have seen their MCP server interactions.
Govern the agent supply chain
MCP servers give agents tools and system access, and a developer can add one in minutes. In a scan of 1,000 MCP servers, 33% carried critical vulnerabilities. Approval alone doesn’t settle the risk either, CVE-2025-54136 showed that a tool definition approved once can be silently changed afterwards. Every MCP server and plugin is governed on an allow-or-deny-by-default basis, and Atlas AIDR keeps monitoring each one for behavior change once it has been approved.
Control what goes in, and what it costs
On routed traffic, Atlas AIDR detects secrets in every prompt and can intervene before the model sees it. It screens the request path for prompt injection and enforces block rules and budgets, along with any business-logic rules you define. Token spend is tracked per request and rolled up by user and team, with a breakout by model as well. Spend doubles as a security signal, since a hijacked or looping agent shows up in token burn as an early warning signal.
See Atlas AIDR in action
Here is what your security team sees in Atlas, from the first captured session to a SOC finding.
Built into MDR, investigated 24/7
When a detection fires on AI telemetry, eSentire investigates and responds. High-severity findings land in the same Atlas platform as the rest of your estate, correlated with endpoint and network telemetry. Our Threat Response Unit hunts proactively, before the attacker moves.
That investigation is included, with no extra modules to buy. Your AI sessions are watched by the same Atlas AI and 24/7 SOC that already protect you across everything else you run.
Atlas AIDR covers Anthropic, OpenAI, Google and Amazon Bedrock as model providers, with native depth for Claude Code, Cursor, GitHub Copilot, Codex and Gemini CLI. No rip and replace required.
Getting started
Atlas AIDR rolls out by policy through the tooling you already use, pointing each AI application at Atlas. Your first captured session arrives within minutes, and developers keep working exactly as they do today.
Your data stays in Atlas. On routed traffic, secrtes are identified before the model sees it and before it is written to storage. Retention follows your policy, and access is role-scoped and governed by SSO.
Atlas AIDR is licensed per user, as a cost-effective, uplift to eSentire Atlas Packages . The fastest way to see what it would show you is a short working session against your own AI-enabled workforce: your AI inventory, the MCP servers your agents use and your AI spend by team.
Assume your agents are already doing things you can’t account for. Talk to our team to see what your AI is doing, and what to do about it.
To learn how eSentire can help you find exposures and defend your organization, connect with an eSentire Security Specialist now.
GET STARTEDABOUT THE AUTHOR
Justin Bailey Senior Director, Product Marketing
Justin Bailey is Senior Director of Product Marketing at eSentire, where he leads go-to-market strategy for eSentire's portfolio spanning MDR, offensive security, and threat intelligence. With deep experience across multiple security disciplines, and intelligence-driven security programs, Justin specializes in translating complex security capabilities into impactful and easy to understand narratives. He works at the intersection of product, marketing, and sales to drive growth through go-to-market activities.