SoftTech AI Registry
Datadog · Datadog MCP Server

Sheet management tools

Datadog MCP Server sheets toolset through which an external AI application can create, update, inspect and permanently delete Datadog spreadsheets, including their tables, sheets and pivots.

Recorded characteristics

Function
Architecture: user → external AI application/model → the AI reasons and selects a Sheets MCP tool → Datadog MCP Server receives the tool call → Datadog performs the authorised operation → the spreadsheet changes. Datadog MCP Server is not the reasoning agent. Boundary: the Datadog MCP Server `sheets` toolset only. Excludes Forms (#145), Workflows (#144), Dashboards, generic data-query MCP tools, and functionality available only through the ordinary Sheets UI. Tool inventory (6): - upsert_datadog_spreadsheet — "Creates or updates a Datadog spreadsheet's tables, sheets, and pivots in a single call." Requires Sheets Read and Sheets Write (plus MCP write permission). Exact create-versus-update semantics are not established. Datadog examples (intended use only, not schema evidence): a logs table with service, status, host, timestamp and message columns filtered to errors; a monthly cloud-spend spreadsheet by provider and service with month-over-month percentage change. - get_datadog_spreadsheet_reference — read-only; "Returns reference documentation for building inputs to upsert_datadog_spreadsheet." Its example questions (log-import tabs, cell formatting) do not by themselves prove upsert supports those settings. - search_datadog_spreadsheets — read-only; searches spreadsheets by name or owner; returns a paginated list with IDs and names. - get_datadog_spreadsheet — read-only; retrieves a spreadsheet by ID, returning tables (tables[].id), pivots (pivots[].id) and sheets (sheets[].id) "with their configurations". - get_datadog_spreadsheet_tab_data — read-only; retrieves paginated data from a table, sheet or legacy pivot tab. - delete_datadog_spreadsheet — "Permanently deletes a Datadog spreadsheet by ID. This action cannot be undone." Requires Sheets Write (plus MCP write permission). Persistent objects established: the spreadsheet, its tables, sheets and pivots, and their IDs/configurations. Not claimed: exact support for formulas, calculated columns, formatting, filters, specific data-source configuration or pivot dimensions/calculations. Action-taking is established independently by spreadsheet creation, update and permanent deletion.
Data access
Data access: spreadsheet metadata and configuration; tab data from tables, sheets and legacy pivot tabs. Tab data requires Sheets Read AND the read permission for the underlying data source (Datadog's example: Logs Read Data for log-backed tables) — two separate permission layers: spreadsheet access and underlying-data access. Source-data boundary: Sheets tables query live Datadog data sources (for example logs, RUM, cloud cost). No source-data modification is established for this MCP capability. External action (no): every positively established persistent effect terminates in Datadog Sheets. No established Sheets MCP operation writes back to an outside source, modifies source data, triggers a workflow or invokes an external system. External action is not inferred from live-data querying. Data refresh: Sheets tables query live data; whether a spreadsheet stores query results or re-fetches underlying data in every relevant circumstance is not publicly established.
Actions
Can take actions
External actions
No
Human confirmation
Not established
Permission basis
Mixed
Administrative control
Authentication: Datadog MCP Server-wide methods — OAuth, Personal Access Token, Service Access Token, API + application keys. No Sheets-specific authentication method is established. MCP permissions: mcp_read / mcp_write, kept separate from Sheets permissions. Sheets permissions: Sheets Read (sheets_read, "View Sheets", in the Datadog Read Only Role); Sheets Write (sheets_write, "Create and change Sheets", in the Datadog Standard Role). Permission basis (mixed): Datadog's MCP Server-wide authentication positively supports (A) user-associated authority and (B) service-account authority, and no evidence makes Sheets an exception. This is not mixed because of uncertainty. Narrower unresolved point: Datadog does not repeat the authority model on the Sheets tool entries. Ownership: ownership of an MCP-created spreadsheet is not publicly established; search can filter by owner. No inference about which principal becomes owner. Sheets access (product-level, separate from MCP authentication, Sheets Read/Write and source-data permissions): by default all users have full access to spreadsheets; editors can restrict edit access to selected roles, teams or users, with others getting Viewer. The Sheets MCP toolset is not claimed to configure sharing, ownership, editor restrictions, public links or collaborators. Human confirmation (not_established): no Datadog-enforced confirmation for create, update or delete. Irreversible deletion does not itself establish confirmation; #144's workflow-delete confirmation and MCP-client approval are not inherited. Deletion/recovery: deletion is permanent and cannot be undone per Datadog. No trash, restore, rollback, version history or undo was established in the investigated evidence; this is not a claim that no Datadog recovery mechanism exists anywhere. Audit Trail: Datadog states all MCP tool calls are recorded in Audit Trail as MCP actions, with tool name, arguments, user identity and MCP client. No separate spreadsheet-version auditing is inferred. Tool exposure: default MCP toolset is core; sheets is not part of it and must be requested (toolsets=sheets); toolsets=all includes generally available toolsets including sheets; omit_tools can remove individual tools (no Sheets-specific example).
Default state
Conditional
Availability
Default conditional: requires an external MCP-capable AI application, a Datadog MCP Server connection, the sheets toolset requested, authentication, MCP permissions, Sheets permissions, and underlying-data permission where tab data is read. Lifecycle: active. Datadog explicitly lists `sheets` among generally available MCP toolsets ("These toolsets are generally available") — stronger release-status evidence than #145 Forms. Relationship to #144 (Workflow management and execution tools): separate — #144 persists workflows, can execute them and has external effect yes; #147 persists analytical spreadsheet structures, no workflow execution, external effect no; different toolset and permission family. Relationship to #145 (Form management tools): separate — #145 persists a data-collection form plus linked response datastore; #147 persists analytical spreadsheet structures over Datadog data; different toolset, purpose, structure, permissions, data relationship and deletion behaviour. Other Registry spreadsheet records (Google Workspace/Gemini Sheets actions, Smartsheet, Zapier) are different products and architectures; "spreadsheet" alone is not duplication.
Licensing
Requires Datadog Sheets access and the Datadog MCP Server; no separate MCP-specific licence was established.
External model or provider
No fixed model or provider. The external AI application supplies the reasoning model.
Limitations and uncertainty
1. Exact upsert semantics (create versus update) not established. 2. Exact supported configuration fields not established. 3. Omitted-field behaviour during update not established. 4. Version history not established. 5. Exact table configuration support (data source, filters, columns, calculated columns) not established. 6. Exact sheet/formula/formatting support not established. 7. Exact pivot configuration support not established. 8. Whether upsert can reference data the caller cannot read not established. 9. Ownership of an MCP-created spreadsheet not established. 10. Delete confirmation/recovery beyond documented irreversibility not established. 11. Datadog does not restate the server-wide authority model on the Sheets tool entries. 12. Exact live-data refresh/storage behaviour not established. 13. MCP management of sharing/access settings not established. 14. No Sheets API reference page exists; operational detail is limited to one-line tool descriptions. 15. No demonstrated end-to-end run; Datadog states the MCP Server is under significant development.

Evidence