Solutions / Shadow and zombie API discovery
You can't defend what you can't see.
Undocumented, deprecated, and forgotten endpoints are where breaches start. Cerberus discovers every API and agent on your surface from real traffic, not a stale spec.
inventoryrisk-rankmonitorIllustrative product interface. The figures shown are an example of how Cerberus presents a detection, not benchmark or performance results.
Drift from spec
What ships rarely matches the OpenAPI doc it was supposed to follow.
Deprecated, still live
Old versions stay reachable for years after they leave the docs.
Agents add surface
Every new MCP server and tool is fresh, unmonitored attack surface.
How it works
An inventory built from traffic.
Specs go stale the day they are written. Real traffic does not.
Coverage
What Cerberus catches here.
Shadow APIs
Live endpoints that never made it into any documentation.
Zombie endpoints
Deprecated versions still reachable and still exploitable.
Undocumented MCP servers
Agent tool endpoints stood up outside any review.
Untracked agents
Autonomous actors operating without an owner or a policy.
Spec drift
Routes and parameters that diverge from the published contract.
Sensitive data exposure
Forgotten endpoints quietly returning PII or secrets.
FAQ
Common questions
What is a shadow API?
A shadow API is a live endpoint that never made it into your documentation or inventory. It usually ships as a debug route, an internal helper, or a quick integration, then stays reachable in production. Because nothing tracks it, it gets no review, no monitoring, and no patching, which makes it a favorite entry point for attackers.
What is a zombie API?
A zombie API is a deprecated endpoint that is still reachable after its replacement shipped. The docs say v3, but the traffic says v0 still answers. Old versions miss the security fixes and validation the current version gets, so they stay exploitable for years unless something is watching for them.
How does Cerberus discover endpoints without a spec?
Cerberus builds the inventory from real traffic. Because the instrumentation runs inside your application, every request that reaches your code is observed, including routes no spec mentions. Each discovered endpoint is classified as documented, shadow, or zombie, risk-ranked by exposure and sensitive-data flow, and monitored from the moment it first appears.
Does discovery cover AI agents and MCP servers?
Yes. Agents and MCP tool endpoints are discovered the same way, from the traffic they actually generate. New tool servers and autonomous actors show up in the inventory alongside your APIs, each with an owner, a history, and the same detection pipeline watching them from day one.
See the whole surface.
See Cerberus read your own traffic, human and agentic, in one walkthrough tailored to your stack.
