Every enterprise IT leader we talk to has the same uneasy feeling: they know their developers are building AI agents, but they don’t know where, how many, or what those agents can actually do. It’s happening right now in your organization.
New shadow IT is here, and the stakes are higher
A decade ago, shadow IT meant employees signing up for Dropbox because the corporate file server was too slow. IT teams scrambled to build policies, provision approved tools, and recover visibility they’d lost.
AI agents are following the exact same pattern only faster, and with consequences that go beyond compliance headaches. Shadow IT stored files in the wrong place, Shadow AI runs autonomous systems outside your visibility, with no circuit breaker when something goes wrong, and no record of what they do.
Agent Scanners already discovers agents from Amazon Bedrock, Amazon AgentCore, Azure AI Foundry, Microsoft Copilot Studio, Google Vertex AI, and GoDaddy – bringing them into a unified registry where they can be governed and tracked. Today, we are extending that coverage to four more platforms where enterprise agents are being built.
Your data science team is building agents in Databricks. Your analytics engineers are building AI workflows in Snowflake Cortex. Your LLM developers are orchestrating agents in LangChain and tracking them in LangSmith. And some teams are running Claude managed agents through the Claude API.
None of these agents show up in your IT inventory. None of them are governed. None of them have been assessed for what data they can access, what actions they can take, or what happens when they fail.
One registry for every platform your agents call home
In January, we launched Agent Scanners and since then, we continue to expand the number of platforms supported. Today, we’re adding Databricks, Snowflake, LangSmith, and Claude. The goal isn’t just coverage for its own sake. It’s closing the gap between the agents your teams are building and the agents your organization actually knows about. Here’s what each new scanner surfaces and why it matters for governance:
Databricks
Data-science ML teams operationalize agents at scale using ML models and notebooks. Agent Scanners connects to Databricks Unity Catalog and surfaces every agent via the model serving endpoint in your environment giving you the first complete picture of what’s running in your data platform and what it can reach.
Snowflake
Snowflake Cortex Agents sit directly on top of your enterprise data, with access to the datasets your business runs on. Agent Scanners pulls from Snowflake’s native Cortex Agents API, surfacing every agent so your data governance team knows exactly which agents are operating on which datasets, not just which humans
LangSmith
LangChain is one of the most widely adopted agent frameworks in the developer community. LangSmith is where developers trace, evaluate, and deploy agent workflows. If your engineering teams are building with LangChain, their agents are in LangSmith. Now they’re in the Agent registry too, with the same governance posture as every other platform.
Claude
Organizations are building and managing fleets of agents on top of foundation models through the Claude API. Agent Scanners surfaces Claude managed agents into the registry so they’re subject to the same policy and governance as agents from any other platform.
From invisible to governed
For teams new to Agent Fabric, here’s what happens once scanner brings agents into the registry. Agent Fabric enriches each discovered agent by surfacing metadata about its capabilities, data access, autonomy level, and deployment context.
From there, your teams can:
- Apply governance policies and assign ownership
- Evaluate agent behavior against declared capabilities
- Track trust posture as agent configuration changes over time
Review agents before importing
Not every discovered agent should land in the registry automatically. Before an agent is imported, Agent Fabric surfaces it for human review: inspect its metadata, assess its relevance, and make an explicit decision to include or skip it. Your registry reflects intentional governance – not everything that was found.
What’s next?
Every agent in your registry is an agent you can govern, audit, and act on. Every agent outside it is a liability you don’t know you have. With these four scanners, the gap between the two gets meaningfully smaller and the answer to “Do we know what’s running?” extends to four more corners of your enterprise AI stack.
See the list of all available Agent Scanners, and learn more about Agent Fabric to stay informed.