Ask a CISO how many AI agents run inside their company and you’ll get a confident answer that’s wrong by an order of magnitude. Agents arrive as npm packages and pip installs, not procurement tickets.
Scanning for what agents leave behind
Our discovery sensor takes the EDR approach: instead of asking workloads to self-report, it scans hosts for the artifacts agentic software can’t avoid leaving — framework packages, MCP server manifests, local model runtimes, and tool configurations.
$ visiq-discovery scan --output json | jq '.findings[].kind'
"langchain"
"mcp-server"
"ollama-model"
"openclaw-tools"The sensor ships as a standalone binary, beacons its findings to the control plane, and the whole loop is exercised end-to-end in CI against real EC2 instances — because a discovery pipeline you’ve only tested with mocks is a discovery pipeline you don’t actually have.