Welcome to Cerbera AI
Cerbera AI gives security teams visibility and control over how AI is used inside the organization. A lightweight local proxy, deployed through your MDM, inspects traffic to AI tools wherever it originates: a browser tab on ChatGPT, a desktop app like Claude, a coding agent like Claude Code, or an MCP server running locally.
Because it sits in the request path, Cerbera AI covers the entire AI attack surface on a device from a single deployment, and it is privacy by design: by default only metrics leave the device, never prompts or responses.
Key Capabilities
AI Discovery
Inventory every AI tool, model, agent, and MCP server in use across your fleet, then decide what to allow or block.
Rules
A firewall-style catalog with three actions: monitor, redact secrets, or block, scoped by model, device, and behavior.
Agent Controls
Block risky agent behavior such as SSH commands and .env access, and detect personal account usage on Claude and ChatGPT.
MCP Governance
Discover, allow, and block the MCP servers your employees connect to, and connect AI clients to Cerbera.
Privacy by Design
Metrics only by default. Logging prompts requires a deliberate, auditable double opt-in and can be disabled organization-wide.
Openness & Interoperability
Export alerts over OpenTelemetry, drive configuration from the API and MCP, and react to events with webhooks.
Why Choose Cerbera AI?
- Full attack surface: One proxy covers browser LLMs, desktop apps, local coding agents, and MCP, not just one channel.
- Purpose-built for AI: Unlike an EDR (system calls) or a firewall (network), Cerbera AI sees prompts, redacts secrets, and governs AI-specific behavior.
- Privacy by design: The default posture collects only metrics. Seeing a prompt is a deliberate, auditable exception, not the norm.
- Managed model: Cerbera maintains the detection catalog and proposes rules as new tools appear, so your team carries minimal load.
- Invisible in use: Sub-millisecond median latency means the proxy is transparent to users.
