Skip to Content

MCP Server

MCP prompts and resources

Start a guided task with a prompt, and know which reference guides the server gives your AI app.

Prompts

A prompt is a ready set of instructions that your AI app follows step by step. In most apps you pick it from the prompt or slash-command menu. You can also describe the task in your own words, and the app finds the tools.

Prompts take no arguments. The app asks you for what it needs, for example the data app or the client ID.

PromptWhat it does
embed-existing-dashboardEmbed a dashboard you already built: pick the data app, make sure it has an API token, pick the dashboard, create the embed
embed-blank-dashboardGive each client their own dashboard with create_embed and a clientId
brand-my-embedChange colors, fonts, chart styles and card styling to match your brand
query-dataAsk questions in plain words, get rows and a chart, and improve the semantic layer
populate-semantic-layerGenerate table and column descriptions for a datamart, so plain-word questions work better
customize-customer-dashboardAdd, change or remove widgets on one client’s dashboard. Other clients do not change
sql-to-metric-migrationTurn tested SQL into a metric on a client dashboard: check, dry run, then apply with APPLY_TO_PRODUCTION
sql-to-workspace-metricTurn tested SQL into an unpublished workspace metric, with no client ID: check, dry run, then apply

A prompt does not give the app any extra access. Each tool it calls still needs your role and, on the hosted connector, the access you approved.

Resources

The server also gives your AI app these reference guides. The app reads them when it needs a fact, for example a web component attribute.

ResourceWhat it holds
databrain://getting-startedThe object model, an onboarding checklist and common mistakes
databrain://api-referenceA summary of the main API endpoints
databrain://embedding-guideEmbed code for React, Next.js, Vue, Angular, Svelte, SolidJS and plain JavaScript
databrain://filter-referenceFilter types and operators for guest tokens, runtime and dashboard filters
databrain://theme-referenceTheme properties and preset themes
databrain://web-component-referenceEvery property of dbn-dashboard and dbn-metric
databrain://self-serve-referenceAccess settings, metric creation modes and runtime permission overrides
databrain://semantic-layer-guideSemantic layer setup for plain-word questions
databrain://multi-tenancy-guideRow-level security with clientId and the client column
databrain://use-case-guidesThe three main use cases, with tool order and prompt names
databrain://permission-schemaEmbed access settings, guest token permissions and which one wins

Apps that show MCP Apps also get two tile views, ui://databrain/chart-tile and ui://databrain/kpi-tile. They draw the chart and KPI results inside the chat.

See the tools reference for what each tool does.