Continuous AI security · Swiss-built
Across browsers and managed endpoints, Numyra keeps a live inventory of the AI services, agents, and components in use. It scores their risk and keeps the evidence. It enforces your rules before sensitive data reaches an unapproved service or an agent completes an unauthorized action. Nobody slows down.
200+ enterprise workstations already deployed
Install in minutes · Runs alongside your existing security tools
| Severity | Alert | Category | Channel | Status |
|---|---|---|---|---|
| High | Sensitive information processed by an agent | Data exposure | Agent | Blocked |
| Medium | Unapproved generative AI service usage | Shadow AI | Browser | Alert |
| Low | ChatGPT accessed with a free account | Shadow AI | Browser | Alert |
| High | Agent access to a restricted file | Access control | Agent | Blocked |
Every company we speak to tells us the same thing: "Our people are using AI but we do not know which tools."
That is not a security failure. AI adoption is moving faster than governance.
AI services, accounts, coding agents, MCP servers, plugins, and dependencies do not appear in one current view. When a component is vulnerable or unapproved, teams need to know where it runs.
Agents can make mistakes even when no one attacks them. They can misread an instruction, act on stale context, or take an unexpected shortcut. Without operational controls, a mistake can expose data, change systems, or disrupt a critical workflow before security can respond.
Native logs vary by tool and are difficult to manage centrally. When a board, auditor, or security review asks what was in use, where, and when, the required record may not exist.
AI adoption should save time, not create investigations later. Without visibility, security teams have to find affected endpoints, reconstruct agent actions, and gather evidence after the fact.
See the AI services, accounts, agents, and components observed across managed endpoints, where they were found, and how they changed. When a vulnerable component is disclosed, find the affected endpoints from the inventory instead of surveying every team.
Bans push AI use toward personal accounts you cannot see. Set targeted runtime policies instead, for sensitive data and agent actions. Raise an alert or block before an unauthorized action completes. The rules are deterministic: the same situation produces the same decision, with no model deciding on your behalf.
Keep a historical record of observed assets, versions, changes, and the facts behind each risk assessment. When the question arrives from the board, an auditor, or a customer's security review, the technical evidence is already available.
Not a one-time scan
Your AI use changes every week. These five steps repeat, so the picture never goes stale.
Discover AI tools, systems, and activity on endpoints and in the browser.
Create and continuously maintain a technical inventory.
Evaluate the risk of the services, accounts, and agents in use.
Apply deterministic policies that alert or block while AI is working.
Produce technical evidence of the AI activity observed and the controls applied.
Repeats continuously
+ many more, discovered continuously
Three plans, scoped and quoted to your environment.
| Capability | Basicsee | Procontrol | Enterpriseown |
|---|---|---|---|
| Live inventory of the AI services, accounts, and agents in use | |||
| Risk scores for services, accounts, and coding agents | |||
| Audit history retention | 7 days | 90 days | Custom |
| Runtime security policies to alert or block AI use and agent actions | |||
| Deployment on-prem | |||
| SLA and dedicated support |
We scope the right plan and quote it with you.
Installation takes only a few minutes. Numyra configures itself and organizations can deploy it through their existing endpoint management tool. No action required from the end user.
No; Numyra adds AI-specific visibility and risk context that complement existing endpoint security tools.
No; Numyra observes AI activity on the endpoint itself. It does not require TLS interception or custom certificates.
Yes. With Basic and Pro, data can be stored in the EU or Switzerland, based on your data-residency requirements. Enterprise also supports on-premises deployment and customer-controlled storage, including control over data location, access, storage, and retention.
No. Numyra maintains an inventory of approved and unapproved AI services, accounts, agents, and components observed across managed endpoints. This gives security teams a common factual basis for deciding what requires review.
Yes. Numyra can identify account types used with services such as ChatGPT and Claude, including Free, Team, Business, and Enterprise accounts, where identifiable. This information is available in the inventory and can be used to define policies that alert on or block access to non-corporate accounts.
Yes. The Basic plan provides a live inventory, risk context, and audit history. Runtime policies and blocking are available with Pro and Enterprise, so teams can establish visibility before introducing controls.
Security teams can identify the managed endpoints where a component or version was observed and review its history. This speeds up scoping, supports targeted containment, and helps limit the potential blast radius without surveying every team manually.
Numyra combines an endpoint component with a browser extension to observe AI activity across managed devices and the browser. Together, they provide a continuously updated view of browser-based GenAI services, desktop AI applications, coding agents, and associated components.
