AI agents can now manage n8n data tables via MCP

n8n workflows can store structured data in tables, but that data has historically been accessible only through the platform UI or internal APIs. This PR exposes data table operations to external callers — specifically AI agents using the Model Context Protocol — through seven new tools.

AI agents can now search for tables by name or project, create new tables with specific column schemas, rename tables, add or delete columns, and insert rows. Each operation runs through a new proxy layer that enforces user-scoped permissions: before any write action, the system verifies the calling user has the appropriate scope on the target project. Read operations pass through the same permission check. Protected instances — those in source control read-only mode — block all write operations with a clear error message.

The tools use a shared schema library for common types like column names, data types, and pagination limits. The implementation follows existing MCP patterns in the codebase, with consistent telemetry tracking for each tool invocation.

This sits within a broader effort to expand MCP tool coverage. The proxy architecture ensures permissions are enforced consistently as new tools are added.