Skip to content
dinobase

Mastra

Dinobase integrates with Mastra via MCP. Mastra has first-class MCP support, so Dinobase’s MCP server works with zero adapter code.

Install

Section titled “Install”
Terminal window
npm install @mastra/core @mastra/mcp @ai-sdk/anthropic zod
pip install dinobase

Set up your data sources:

Terminal window
dinobase init
dinobase add stripe --api-key sk_test_...
dinobase add hubspot --api-key pat-...
dinobase sync

See Connecting Sources for the full list of 100+ supported sources, and Syncing & Scheduling for background sync options.

Quick Start

Section titled “Quick Start”
import { Agent } from "@mastra/core";
import { MCPClient } from "@mastra/mcp";
import { anthropic } from "@ai-sdk/anthropic";


// Connect to Dinobase MCP server
const mcp = new MCPClient({
  id: "dinobase-mcp",
  servers: {
    dinobase: {
      command: "dinobase",
      args: ["serve"],
    },
  },
});


// All 7 Dinobase tools discovered automatically
const agent = new Agent({
  id: "data-analyst",
  name: "Data Analyst",
  instructions: "You are a data analyst. Query business data via SQL.",
  model: anthropic("claude-sonnet-4-6"),
  tools: await mcp.listTools(),
});


const response = await agent.generate(
  "Which customers have overdue invoices?"
);
console.log(response.text);


await mcp.disconnect();

How It Works

Section titled “How It Works”

Mastra’s MCPClient spawns the Dinobase MCP server as a subprocess (dinobase serve) and communicates via stdio. All 7 Dinobase MCP tools are discovered automatically and made available to the agent.

This is the same MCP server used by Claude Desktop and Vercel AI SDK — same tools, same data, different framework.

Available Tools

Section titled “Available Tools”

Via MCP, the agent gets access to:

ToolDescription
queryExecute SQL queries (DuckDB dialect)
describeGet table schema, types, and sample data
list_sourcesList connected sources with freshness status
refreshRe-sync a stale data source
confirmExecute a pending mutation (write-back)
confirm_batchExecute multiple pending mutations
cancelCancel a pending mutation

Custom Tools (Alternative)

Section titled “Custom Tools (Alternative)”

If you want more control over tool behavior, you can wrap the Dinobase CLI using createTool():

import { createTool } from "@mastra/core/tools";
import { z } from "zod";
import { execSync } from "child_process";


export const dinobaseQuery = createTool({
  id: "dinobase-query",
  description: "Execute SQL against Dinobase",
  inputSchema: z.object({
    sql: z.string().describe("SQL query to execute"),
  }),
  outputSchema: z.object({ result: z.string() }),
  execute: async ({ sql }) => ({
    result: execSync(`dinobase query ${JSON.stringify(sql)}`, {
      encoding: "utf-8",
    }),
  }),
});

See examples/tools.ts for all four tool wrappers.

Background Sync

Section titled “Background Sync”

Keep data fresh while the MCP server runs:

const mcp = new MCPClient({
  id: "dinobase-mcp",
  servers: {
    dinobase: {
      command: "dinobase",
      args: ["serve", "--sync", "--sync-interval", "30m"],
    },
  },
});

Next steps

Section titled “Next steps”