What is an MCP Resource?
An MCP resource is a read-only data source exposed by an MCP server that provides context to AI agents — such as files, database records, API responses, or configuration — without triggering side effects.
WHY IT MATTERS
While MCP tools are actions (they do things), MCP resources are data (they provide information). Resources give agents context without executing operations — reading a file, viewing a database schema, checking a configuration value.
Resources can be static (a fixed configuration file) or dynamic (current system state, live metrics). They support URI-based addressing and can include metadata about content type and freshness.
Even read-only data can be sensitive. Source code, environment variables, credentials files, customer data — all might be exposed as MCP resources. Access controls on resource reads are just as important as controls on tool calls.
HOW POLICYLAYER USES THIS
PolicyLayer can enforce access controls on MCP resource reads. YAML policies can restrict which resources an agent is permitted to access — for example, allowing reads from config:// URIs but denying access to secrets:// URIs. This prevents agents from accessing sensitive data exposed by the server.