Skip to main content

Getting Started with MCP Tools & Servers

This guide walks you through the end-to-end workflow for turning a Nexset into an AI-ready MCP tool, deploying it as an MCP Server, and connecting an AI agent. By the end, you will have an AI agent that can discover and invoke tools backed by your Nexla data.

Prerequisites​

Before you begin, make sure you have the following:

  • An active Nexla account with at least one active Nexset. If you need to create a Nexset, see the Quick Start Guide.
  • Access to the AI tab in the Nexla Platform. Contact your Nexla administrator if you do not see the AI tab in your navigation.
  • A Nexla service key (required for API access and for configuring MCP clients). See Service Keys for instructions on creating one.
  • An MCP-compatible client for testing, such as Claude Desktop, Cursor, or another agent that supports the Model Context Protocol.

Step 1: Navigate to the AI Tab​

Open the Nexla Platform and click the AI tab in the main navigation bar. This takes you to the MCP Tools & Servers dashboard, which is your central hub for managing tools, ToolSets, MCP Servers, and the MCP Gateway.

The dashboard displays an overview of your existing resources, including active tools, deployed servers, and recent activity.

Step 2: Create a Tool from a Nexset​

Tools are generated directly from your Nexsets. Each tool inherits its schema and behavior from the underlying Nexset.

  1. From the AI tab, navigate to the Nexsets browser.
  2. Locate and select the Nexset you want to expose as an AI tool.
  3. Click Generate Tool to open the Tool Generator wizard.

The wizard guides you through the following stages:

  1. Select Nexset -- Confirm the Nexset you want to generate a tool from. Review the Nexset's schema and sample data.
  2. Review configuration -- Review the generated tool name, description, and input/output schema. You can adjust the tool description to provide better context for AI agents.
  3. Generating -- Nexla automatically generates the tool definition, configures the execution logic, and registers the tool. This step is automatic.
  4. Success -- The tool has been created. You can optionally add it to an existing ToolSet immediately, or do so later.
tip

Write clear, descriptive tool descriptions. AI agents rely on the tool description to decide when and how to use a tool. A well-written description leads to more accurate tool selection by the agent.

After the wizard completes, the tool appears in the Tools Registry with a status of active. You can view and manage all your tools from the Tools Registry page.

Step 3: Create a ToolSet​

A ToolSet groups one or more tools into a deployable collection. You need at least one ToolSet to deploy an MCP Server.

  1. Navigate to the ToolSets page from the AI tab.
  2. Click Create ToolSet.
  3. Enter a name for the ToolSet. Choose a name that reflects the purpose or domain of the tools it contains (e.g., "Sales Data Tools" or "Inventory Lookup").
  4. Optionally, add a description to help collaborators understand what the ToolSet is for.
  5. Select tools to include from the Tools Registry. You can add multiple tools from different Nexsets.
  6. Click Save to create the ToolSet.
note

You can update a ToolSet at any time by adding or removing tools. Changes to a ToolSet are reflected in any MCP Server deployed from it after the server is redeployed or refreshed.

Step 4: Deploy as an MCP Server​

Once your ToolSet is ready, you can deploy it as an MCP Server that AI agents connect to.

  1. Open the ToolSet you want to deploy by clicking on it in the ToolSets list.

  2. From the ToolSet detail page, click Export as MCP Server.

  3. Select the target environment: dev, stage, or prod. Use dev for initial testing.

  4. Nexla provisions the MCP Server and generates a unique MCP URL in the following format:

    https://api-genai.nexla.io/mcp/service_key/{server_key}

  5. Copy the MCP config provided on the confirmation screen. You will use this to configure your AI client in the next step.

warning

Keep your server key secure. Anyone with the server key and a valid service key can access the tools exposed by this MCP Server. Rotate keys immediately if you suspect unauthorized access.

Step 5: Connect an AI Agent​

With your MCP Server deployed, you can now connect an MCP-compatible AI client. The following example demonstrates how to configure Claude Desktop.

Configure Claude Desktop​

Add the following configuration to your Claude Desktop MCP settings file:

{
  "mcpServers": {
    "nexla": {
      "url": "https://api-genai.nexla.io/mcp/service_key/{your_server_key}",
      "headers": {
        "Authorization": "Bearer YOUR_NEXLA_SERVICE_KEY"
      }
    }
  }
}

Replace the placeholder values:

  • {your_server_key} -- The server key from Step 4.
  • YOUR_NEXLA_SERVICE_KEY -- Your Nexla service key. See Service Keys if you have not created one yet.

Verify the Connection​

  1. Restart Claude Desktop (or your MCP client) to load the new configuration.
  2. Open a new conversation and check the tool list. Your Nexla tools should appear among the available tools.
  3. Ask the agent a question that requires data from your Nexset. The agent should automatically select and invoke the appropriate tool.
tip

If your tools do not appear after restarting, double-check the server URL, service key, and JSON syntax in your configuration file. Ensure the MCP Server status is active in the Nexla Platform.

Next Steps​

Now that you have completed the basic workflow, explore these guides to learn more:

  • Managing Tools -- Create additional tools, update tool configurations, and manage tool versions.
  • Managing ToolSets -- Organize tools into ToolSets and manage ToolSet manifests.
  • Connect MCP Clients -- Detailed instructions for connecting various MCP-compatible clients beyond Claude Desktop.
  • MCP Gateway -- Register external MCP servers and bring third-party tools into your ToolSets.
  • Audit & Receipts -- Monitor tool usage and review the audit trail of all tool executions.

For programmatic access, see the MCP Tools & Servers API.