Skip to main content
Nekt’s MCP Server lets AI agents interact with your Catalog data using natural language. When integrated with an agent, it provides structured context and tools for reasoning, querying, and responding based on your data.
The MCP Server is not available on the free plan.
AvailableInfrastructure
Nekt ExpressYesManaged by Nekt, no separate AWS bill
GCPNot applicable
AWSYesRuns in your AWS account (~$34/month in us-east-1)

AWS costs

If your workspace runs on your own AWS account, the following services are used. Costs vary by region — in Northern Virginia (us-east-1), the fixed cost is approximately $34.00/month. On Nekt Express, infrastructure is fully managed by Nekt — no separate bill.

Setup

Prerequisites

Your cloud environment (or Nekt Express workspace) must be on version 2.0.19 or later.

Video tutorial


Step 1: Enable the MCP Server

  1. In Cloud Settings, click Enable MCP Server.
  2. Enter a custom domain you own. This will be the endpoint for your MCP server.

Step 2: Verify the HTTPS certificate

Add the CNAME record from Nekt to your DNS provider to validate the HTTPS certificate. Allow 5–10 minutes for verification.

Step 3: Add the MCP URL to your DNS

Still in your DNS provider, add the provided CNAME record to point your domain to the MCP server URL. Allow an additional 5–10 minutes. Your MCP server is now live.

Step 4: Create tokens

Create tokens to control access to the MCP server. Each token can be scoped to a specific set of tables.

Available tools

The MCP provides three tools for your agent.

ask_a_question

Ask a natural language question about your data. The tool generates an Athena SQL query, which you can then execute using execute_sql_query. Scope is limited to the tables associated with the token.

execute_sql_query

Executes an Athena SQL query against your data lake and returns results.

save_assistant_response

Call this at the end of an agent workflow after the final response is formulated. It appends the response to the session’s history file, making it available as context in subsequent interactions.

Need help?

Contact our support team if you encounter issues during setup.