Getting Started with Actian MCP Server¶

The Actian MCP Server is distributed as Docker container images, with one dedicated image for each supported Actian database. This guide explains the process of starting the server instance and connecting it to an MCP-compatible artificial intelligence (AI) client.

Prerequisites¶

Before you begin, ensure that the following requirements are met:

• Container runtime: Docker or Podman is installed on the host machine
• Database access: There is network connectivity to a supported Actian database (Ingres, HCL Informix®, Zen, NoSQL, or Analytics Engine)
• AI client: An MCP-compatible client, such as Claude Desktop, Cursor, GitHub Copilot, or Codex already exists

Step 1: Select a Database Image¶

Each Actian database uses a specific container image that is preconfigured with the required drivers. Identify the correct image for the environment:

Step 2: Create a Configuration File¶

For most databases, you need to create a conf.json file that contains the specific connection details.

Each database has unique settings. For more information, see the database configuration document:

• Ingres Configuration
• HCL Informix® Configuration
• Zen Configuration
• NoSQL Configuration
• Analytics Engine Configuration

All database configurations except NoSQL share the following standard MCP server fields:

Step 3: Start the Container¶

To start the server, run the container and mount the configuration file to /app/conf.json. Use the :ro (read-only) flag in the mount command to ensure that the configuration file cannot be modified from inside the container.

Run the following command (replace the image name with the database image as per Step 1):

docker run -d \
    -v $(pwd)/conf.json:/app/conf.json:ro \
    -p 8000:8000 \
    --name=actian-mcp \
    actian/analytics-engine-mcp-server

Step 4: Connect to an MCP Client¶

The server operates in HTTP transport mode. Configure the AI client to connect to the following endpoint:

http://localhost:<port>/mcp

For configuration examples for Claude Desktop, Cursor, GitHub Copilot, fast-agent, and Codex, see Connecting MCP Clients.

Step 5: Verify the Connection¶

Follow these steps to confirm that the server is running and successfully communicating with the database and AI client:

1. Verify the container status

Run the following command to ensure that the container is actively running:

docker ps --filter "name=actian-mcp"

Confirm that the container status is Up.

2. Verify the endpoint

Ping the server to confirm that it is listening for requests:

curl -i http://localhost:<port>/mcp

If the server is ready, it returns a 200 or 307 status code instead of a connection refused error.

3. Test the client integration

Open the configured MCP client. It automatically detects the Actian MCP Server and displays its available tools. Prompt the AI with a standard database request, such as:

"List all tables in the database"

The client will invoke the server's list_tables tool. If the AI client returns a list of the database tables, the end-to-end connection is working correctly.

For the complete list of available tools for each database, see database-specific documentation:

• Ingres tools
• HCL Informix® tools
• Zen tools
• NoSQL tools
• Analytics Engine tools

Step 6 (Optional): Add Authentication¶

If you are deploying the server outside a secure, trusted local environment, it is recommended to use OAuth 2.0 authentication. You can secure the server using an external identity provider like Keycloak or Auth0. See the following authentication documentation for detailed setup instructions:

• Authentication overview
• Keycloak setup
• Auth0 setup

Next Steps¶