HomeBlogmcp-tools-for-ai-agents

mcp-tools-for-ai-agents

Baseil TeamInvalid Date2 min read

title: "Using Baseil's MCP Tools with AI Agents" description: "How to connect your AI agents to any database through Baseil's automatically generated MCP tools." date: "2026-02-14" author: "Baseil Team" tags: ["mcp", "agents", "integration", "workflow"]

Why MCP?

The Model Context Protocol (MCP) is becoming the standard way AI agents interact with tools and data sources. Baseil automatically exposes every connected database as MCP tools — no custom integration code needed.

How It Works

When you connect a database to Baseil, it automatically generates MCP tool definitions:

{
  "tools": [
    {
      "name": "query_customers_db",
      "description": "Query the customers database using natural language",
      "parameters": {
        "query": "Natural language question about customer data"
      }
    },
    {
      "name": "query_analytics_db",
      "description": "Query the analytics database using natural language",
      "parameters": {
        "query": "Natural language question about analytics data"
      }
    }
  ]
}

Your AI agents can now query any database just by calling these tools.

Setting Up Agent Access

1. Start Baseil with MCP Server

baseil serve --mcp --port 3100

2. Point Your Agent to Baseil

Configure your AI agent's MCP client to connect:

{
  "mcp_servers": [
    {
      "name": "baseil",
      "url": "http://localhost:3100/mcp"
    }
  ]
}

3. Let Your Agent Query

Your agent can now make calls like:

Tool: query_customers_db
Input: { "query": "How many new signups this week?" }

And get structured results back instantly.

Cross-Database Queries

The real power comes when agents need data from multiple sources. Instead of building custom connectors for each database, your agent just asks:

Tool: query_all
Input: { "query": "Compare customer growth in our PostgreSQL CRM with event data in our analytics DB" }

Baseil handles the cross-database join automatically.

Workflow: Customer Support Agent

Here's a practical workflow for a customer support AI agent:

  1. Customer asks: "Why was I charged twice?"
  2. Agent calls Baseil: query_billing_db → "Show recent charges for customer ID 12345"
  3. Agent calls Baseil: query_payments_db → "Show payment processor logs for customer 12345 in the last 7 days"
  4. Agent synthesizes both results and responds to the customer

No custom code. No database drivers. Just natural language through MCP tools.

Best Practices

  • Scope your tools — expose only the databases each agent needs
  • Use rules — pre-define business terms so agents get consistent results
  • Monitor usage — Baseil logs every query for audit trails
  • Cache hot queries — pin frequently-asked agent queries for instant responses