> ## Documentation Index
> Fetch the complete documentation index at: https://docs.plumi.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Studio

> Build, test, and manage MCP tools for AI assistant integrations

## Overview

MCP Studio is your integrated development environment for creating Model Context Protocol (MCP) tools. Build custom tools that let AI assistants interact with your data and systems.

## Features

MCP Studio includes four main areas:

<CardGroup cols={2}>
  <Card title="Tool Builder" icon="wrench">
    Create and configure MCP tools with a visual editor.
  </Card>

  <Card title="Testing Playground" icon="flask">
    Test your tools before deploying to production.
  </Card>

  <Card title="Versions" icon="code-branch">
    Manage tool versions and rollbacks.
  </Card>

  <Card title="Git Integration" icon="git-alt">
    Connect to GitHub for version control.
  </Card>
</CardGroup>

## Tool Builder

The Tool Builder provides a visual interface for creating MCP tools:

### Creating a Tool

1. Navigate to **MCP Studio > Tool Builder**
2. Click **New Tool**
3. Define tool properties:
   * **Name**: Unique identifier for the tool
   * **Description**: What the tool does (shown to AI)
   * **Parameters**: Input parameters the tool accepts
   * **SQL Query**: The query to execute

### Tool Parameters

Define parameters that AI assistants can pass to your tool:

```json theme={null}
{
  "name": "customer_id",
  "type": "string",
  "description": "The unique customer identifier",
  "required": true
}
```

### Example Tool

```yaml theme={null}
name: get_customer_orders
description: Retrieve orders for a specific customer
parameters:
  - name: customer_id
    type: string
    required: true
  - name: limit
    type: integer
    default: 10
query: |
  SELECT * FROM orders
  WHERE customer_id = {{customer_id}}
  LIMIT {{limit}}
```

## Testing Playground

Test your tools before deploying:

1. Go to **MCP Studio > Testing Playground**
2. Select the tool to test
3. Enter parameter values
4. Click **Run** to execute
5. Review the results

<Tip>
  Test with edge cases and invalid inputs to ensure your tools handle errors gracefully.
</Tip>

## Version Management

### Creating Versions

1. Go to **MCP Studio > Versions**
2. Click **Create Version**
3. Add a version name and description
4. Select which tools to include

### Rolling Back

If a new version has issues:

1. Go to **Versions**
2. Find the previous stable version
3. Click **Rollback** to restore it

## Git Integration

Connect your MCP tools to GitHub for version control:

### Connecting GitHub

1. Go to **MCP Studio > Git**
2. Click **Connect GitHub**
3. Authorize Plumi to access your repositories
4. Select or create a repository for your tools

### Syncing Tools

* **Push**: Save your tools to GitHub
* **Pull**: Load tools from GitHub
* **Commit History**: View changes over time

## Best Practices

<Tip>
  **Write clear descriptions**: AI assistants use tool descriptions to decide when to use each tool.
</Tip>

<Tip>
  **Validate inputs**: Add parameter validation to prevent SQL injection and invalid queries.
</Tip>

<Tip>
  **Test thoroughly**: Use the playground to test all parameter combinations before deploying.
</Tip>

## Using Your Tools

Once published, your tools are available via MCP:

```json theme={null}
{
  "mcpServers": {
    "plumi": {
      "url": "https://your-workspace.plumi.ai/sse?api_key=YOUR_KEY"
    }
  }
}
```

AI assistants can then use your custom tools alongside built-in Plumi tools.