1. Determine if a local or remote dbt MCP server is needed.
2. Identify the specific AI client being used (Claude Desktop, Claude Code, Cursor, VS Code).
3. Gather necessary credentials and paths (dbt project directory, dbt executable path, account ID, environment ID, token).
4. Configure the AI client with the appropriate JSON configuration, including environment variables and server settings.
5. Test the local server configuration by running `uvx dbt-mcp`.
6. Verify the connection in the AI client by listing dbt metrics or models.
7. Troubleshoot common errors such as 'uvx not found' or 'Could not connect to MCP server'.
SynapticSkills