AI Discovery

Find shadow AI — unknown agents, unmanaged MCP servers and unauthorized model usage — and bring it under governance.

Almost no enterprise knows how many AI agents it runs. Discovery surfaces the AI nobody registered and lets you promote it to governed status.

What Discovery surfaces#

Ungoverned agents — discovered = true, lifecycleState = DISCOVERED, or authMethod = NONE.
Unapproved MCP servers — approvalStatus = PENDING or trustLevel = UNTRUSTED.
On the roadmap: unknown models, unmanaged secrets, and AI workloads running in Kubernetes, Docker and cloud accounts.

Find shadow AI#

Discovery is derived from the directory and registry. List agents and MCP servers and filter for ungoverned items:

curl "https://your-company.authspoke.com/api/v1/ai/agents" -H "Authorization: Bearer $TOKEN"
curl "https://your-company.authspoke.com/api/v1/ai/mcp-servers" -H "Authorization: Bearer $TOKEN"

The overview returns shadowAgents and shadowMcpServers counts directly.

Promote to governed#

Bring a discovered agent under control by updating its lifecycle:

curl -X PUT "https://your-company.authspoke.com/api/v1/ai/agents/$ID" \
  -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
  -d '{ "name": "repo-janitor", "discovered": false, "lifecycleState": "REGISTERED" }'

Approve a pending MCP server the same way by setting approvalStatus to APPROVED.