Watchtower / shared-data access
Application-to-MCP coverage
This inventory links application workspaces, shared datasets and public reference content to their MCP read paths. It records what was tested and the source limitations that still apply.
Loading the current inventory…
Explore the inventory
| Name / identifier | Access and verification | Read path and qualifications |
|---|
Reading every admitted field
Use watchtower.get_application_coverage for the inventory and watchtower.read_data_route for the listed source. Follow page.nextOffset with the same MCP snapshot for field pages. Source row pagination has its own parameters and continuation.
Public references additionally support view=structure, view=fields and view=search-fields. Select a source JSON Pointer, follow fieldOffset / nextFieldOffset, and retain sourceSnapshot. For a long source string, select that string and continue with textOffset / nextTextOffset. These views can reach a large record without squeezing its entire geometry or text into one response.
Exact large integers and decimals that cannot round-trip through ordinary JSON numbers are returned as explicitly typed strings. Zero, missing values and false remain distinct. Sensitive fields retain the existing redaction policy.
Scope boundaries
Private accounts, saved investigations, portfolios, personal layouts, credentials, billing and operational actions are outside this shared-data inventory. Mixed workspaces retain access to their shared input data. Browser-local calculations depend on their input snapshots and clock; their shared inputs and reference/model definitions are identified separately from stored observations. External iframe, stream and tile representations remain source-controlled references.