dbt Cloud MCP Integration

dbt Cloud is where many data teams define the canonical metrics and models the rest of the business reports against. Connecting dbt Cloud to clariBI lets your AI analyses call the dbt Semantic Layer directly — so when someone asks for MRR or churn, you get the analyst-team-approved definition instead of a re-derived approximation.

Why connect dbt Cloud

Most "what does this number mean?" disputes between teams happen because two systems calculated the same metric two different ways. dbt's Semantic Layer fixes this by making one definition the source of truth.

With dbt Cloud connected, you can ask "What is monthly active users by signup channel last quarter?" or "Show me our cumulative new ARR for FY26 by region", and the AI engine routes through the dbt Semantic Layer — so the number matches what your data team publishes. clariBI also reads model lineage + freshness, so questions like "is this dashboard up to date?" or "what changed upstream?" get accurate answers. The integration is read-only at the catalog level: write CLI tools (build, run, test, trigger_job_run, retry_job_run) are filtered out so clariBI cannot mutate dbt resources.

How the connection works

clariBI talks to dbt Cloud through its hosted MCP server at https://cloud.getdbt.com/api/ai/v1/mcp/. Authentication uses an OAuth flow that clariBI registers itself for (no developer console setup on your side). Tokens stay encrypted server-side and never leave clariBI in clear form.

sequenceDiagram
    actor U as You
    participant C as clariBI
    participant V as dbt Cloud
    U->>C: Click Authorize with dbt Cloud
    C->>V: Open OAuth authorization
    V-->>U: Grant read access?
    U->>V: Approve
    V-->>C: Authorization code
    C->>V: Exchange code for tokens
    V-->>C: Access + refresh tokens
    C->>C: Encrypt and store credentials
    C-->>U: Connection ready

Available tools

The AI engine selects from these read-only dbt Cloud tools during analyses. None of them modify data in dbt Cloud.

Data flow during analysis

When you ask a question that maps to dbt Cloud, the AI engine routes to the right tool, reads the result, and pairs the answer with a chart you can pin to a dashboard.

sequenceDiagram
    actor U as You
    participant C as clariBI
    participant AI as AI engine
    participant V as dbt Cloud
    U->>C: Ask a question about dbt semantic metrics, model lineage, and run history
    C->>AI: Plan the analysis
    AI->>V: Call the right tool
    V-->>AI: Tool result
    AI->>AI: Summarize and chart
    C-->>U: Answer plus visual

Setting up the connection

1. {'title': 'Confirm your dbt plan supports remote MCP', 'description': 'Remote MCP with OAuth is available on dbt Cloud Enterprise and Enterprise+ accounts. Lower-tier accounts can still connect via a Personal Access Token; see the dbt MCP docs for the token flow.'}
2. {'title': 'Copy your account-specific MCP endpoint URL', 'description': 'In dbt Cloud go to Account settings → Access URLs → MCP Endpoint URL. The URL is account-specific because multi-cell accounts use a per-account prefix (e.g. https://your-account.us1.dbt.com/api/ai/v1/mcp/) instead of the default cloud.getdbt.com.'}
3. {'title': 'Paste the URL in the clariBI connect modal', 'description': 'Use the URL you copied. clariBI hands it to the OAuth flow; dbt opens a browser sign-in and shows you the requested scopes (Semantic Layer read + Discovery read).'}
4. {'title': 'Authorize on the dbt consent screen', 'description': 'Review the scopes shown and confirm. clariBI never requests write scopes.'}

Permissions and data access

clariBI requests read-only scopes covering the Semantic Layer (metrics, dimensions, entities, saved queries) and Discovery (model metadata, lineage, freshness, test results). Job-run and project metadata are also read-only. Write CLI tools (build, run, test, clone, trigger_job_run, retry_job_run, cancel_job_run) and direct SQL execution (execute_sql) are filtered out at the catalog layer and never reachable through clariBI. Disconnect anytime from Settings → Integrations or revoke access from Account settings → Integrations → App integrations in dbt Cloud.

Troubleshooting

Related articles

• MCP server integrations overview
• Building dashboards
• Asking questions with AI
• All data source integrations