New: the Agent Incident Ledger: disclosed incidents of AI gone rogueExplore the ledger

EDR sees the process. Not the agent.

EDR watches processes, files and network activity on the machines it sits on. To it, an AI agent is one more process: it sees the agent start, not the session inside it. And the agents that run in someone else’s cloud never touch the endpoint at all.

EDRendpoint detection and response
Manifoldruntime, inside the agent
CoverageWhich agents it sees
Partly covered: Agents running on managed machines, seen as processes.
Covered: Every agent where it runs: endpoints, browsers, VMs and the cloud.
Inside the sessionWhat the agent does after the prompt
Partly covered: The process starting, and its file and network activity. Not the decisions behind them.
Covered: Every tool call, MCP request, file action and command, captured inside the session.
Supply chainMCP servers, skills and plugins
Not covered: Another file on disk or another child process, with no view of what it gives the agent.
Covered: Inventoried per agent, including ones never run, and scored by Manifest.
IdentityWho the agent acted for
Partly covered: The user account on the machine.
Covered: The person the agent acted for, and the account behind it, personal accounts included.
Agents in the cloudPersonal agents, cloud VMs
Not covered: Agents that run in a vendor’s cloud, like personal agents over iMessage or WhatsApp, never reach the endpoint.
Covered: Agent use found across your environment, with what each agent has connected.
ResponsesWhat it can do when something goes wrong
Partly covered: Kill the process or isolate the machine.
Covered: Alert, advise, correct, ask the user, block the call or end the session, inside the harness.

Keep your EDR. Manifold connects the security stack you already have, from identity and EDR to MDM, and adds what happens inside the agent.

See Manifold in action.

A short tour of the platform, from the first agent discovered to the first policy enforced.