Masthead MCP Tools Reference

The Masthead Model Context Protocol server exposes a set of read-only and mutation tools that allow AI agents to interact with your data observability ecosystem in near-realtime.

Below is the reference guide for each tool, including its purpose.

Discovery and context tools

`list_projects`

List all GCP project IDs monitored for your tenant. Use these project IDs as filters, such as projectId, in subsequent incident and table metadata queries.

`list_users`

List all users in your Masthead tenant. This is useful for identifying email addresses before assigning ownership of an incident.

`get_tenant_settings`

Return the tenant settings an agent needs before running FinOps skills: whether the tenant enabled the BigQuery insights export, the export project and dataset (masthead-prod.<dataset>) with its tables (insights, lineage), and the look-back window in days used to flag dead-end and unused tables (30 unless configured otherwise). Skills call this first and stop with a request-access link when the tenant disables the export.

Incident management tools

`list_open_incidents`

List all active unresolved incidents, such as Open, In Progress, or Acknowledged, over the last 14 days, optionally bounded by a limit.

`list_incidents`

Full incident search history with options to filter by project, status, severity, or pipeline type.

`get_incident_details`

Fetch full details and diagnostics for a specific incident group by its UUID.

Incident action and mitigation tools

`bulk_update_incidents`

Update the status, severity, assignee, and notes of one or more incidents in one call. Set only the fields you want to change:

status: the new life-cycle status.
severity: P1–P5, or NONE to clear it.
assigneeEmail: the email of a registered Masthead user, or NONE to unassign. The call fails if no user has that email.
notes with notesMode: ADD, the default, puts the note before the existing notes with a --- line between them. REPLACE overwrites the notes; REPLACE with empty notes clears them.

All changes in a call apply together, and the incident history records each one. The result lists each incident with success, and a message such as Incident not found when an incident wasn’t updated.

Table, BI asset and lineage tools

`get_table_metadata`

Retrieve schema, columns count, isAnomaly status, and alert tier settings for a BigQuery table.

`get_table_pipelines`

Fetch the metadata and status of pipelines writing to a specific table.

`get_table_lineage`

Traces the upstream/downstream tables and pipelines associated with the target table to verify blast radius.

`list_asset_scores`

List the tenant’s scored assets—BigQuery tables and BI assets (Looker dashboards, Looker looks, Looker Studio reports, Google Sheets)—highest score first, with the raw score metrics, the current alert tier, and each asset’s percentile rank within its own asset type in the requested scope, plus p50/p90/p95 per asset type in extra (covering only the asset types present on the page). The score (0-1) ranks an asset’s importance from downstream and upstream compute cost, write, and read frequency, BI assets and service-account readers, normalized within the tenant and asset type. Filter by assetTypes, project, dataset, alertTiers and minPercentile to fetch tier-change candidates; apply with update_table_priority for tables or update_bi_asset_priority for BI assets.

`update_table_priority`

Adjust the monitoring alert level for a BigQuery table.

`update_bi_asset_priority`

Adjust the monitoring alert level for a Looker dashboard, Looker look, Looker Studio report, or Google Sheet, and return that asset’s own score row (matched by uuid) after the change so the caller can confirm the new alertType. The call fails if Masthead didn’t apply the change. Downstream BI asset priority also drives pipeline anomaly alerting.

Data scan tools

See custom data scans for the table columns and grants.

`list_data_scans`

List the tenant’s custom data scans with their table, frequency, processing delay, active flag, and last and next run times. Masthead sets up scans whose source is MANUAL; you can’t change them with these tools.

`create_data_scan`

Validate or create a custom data scan on a BigQuery table. With dryRun set to true, the default, Masthead reads the table as its service account and returns a summary—rows, tables, metrics, time range, and bytes processed—or an error that says what to fix. With dryRun set to false, it also creates the scan and starts its first run right away.

`update_data_scan`

Pause or resume a scan, change its frequency or processing delay, or point it at another table. Changing the table or frequency deletes the scan’s history and analyzes it again from scratch.

`delete_data_scan`

Delete a scan with its stored results and open incidents. The BigQuery table stays in place.