To ensure full transparency, keep the following behaviors in mind when using the Observe AI.

AI Knowledge Graph

The current version of the Knowledge Graph contains all Dataset metadata, such as names and descriptions, as well as a number of rows of example data. This data is available through the AI SRE to all users of the AI SRE. This limitation may make the AI SRE unsuitable for some customers. Per-user feature flags can be enabled, rather than account wide. More restrictive access to the metadata and sample data is under active development.

AI SRE / MCP runtime queries

Runtime queries executed through the MCP Server always use the permissions of the authenticated user or service account.

• If you do not have permission to read a Dataset, such as Kubernetes Logs, then the MCP Server and AI SRE will also be unable to query the Dataset.
• No privilege escalation occurs; the agent can only query Datasets that you can query in Observe UI or through OPAL.

Summary

These two systems behave differently by design:

• Knowledge Graph is a global Dataset catalogue for schema discovery and model reasoning.
• Runtime query execution is strictly governed by your RBAC permissions.