Overview

Foundational provides its customers with an MCP (Model Context Protocol) server, enabling them to leverage the full context around their data stack directly within their preferred AI tools. This context includes end-to-end lineage, usage information for models and dashboards, code definitions for tables, models or upstream operational databases, ownership details, and more, all surfaced directly in your preferred AI tool and powered by Foundational.

This allows customers to easily integrate this rich, accurate information about their data stack into their favorite AI tools—whether using AI coding assistants like Cursor or Claude Code, AI chatbots like Claude Desktop, or even custom internal AI deployments.

What is the MCP Server?

The Model Context Protocol (MCP) Server, introduced by Anthropic, allows AI assistants and other applications to access APIs and contextual data that might otherwise be inaccessible to the core AI model. It acts as a secure gateway or translator. For instance, an MCP server connected to a specific tool (like GitHub) could allow an AI to translate a natural language request ("List my open Pull Requests") into the appropriate API calls for that tool and return the correct information.

Example Use Cases

Use AI to refactor and write data engineering pipelines
When tasked with modifying or creating data pipelines (e.g., dbt models, Airflow DAGs, Spark jobs), AI tools integrated with Foundational MCP can access critical context. Before suggesting code changes, the AI can query the MCP server to understand the upstream sources, downstream dependencies (dashboards, other tables, ML models), column data types, and existing definitions for the tables involved. This helps the AI write safer, more accurate code, preventing breaking changes, ensuring consistency with existing logic, and even suggesting optimizations based on table usage patterns identified by Foundational.
Ask AI questions about your data stack
Instead of manually navigating complex lineage graphs or searching through documentation, you can ask your AI assistant natural language questions about your data. For example: "Where does the fct_user_sessions table get its data from?", "Which dashboards use the dim_products table?", "Who is the owner of the customer_pii_data dataset?", or "Show me the description for the order_status column." The Foundational MCP server provides the AI with the necessary lineage, metadata, and usage information from your Foundational workspace to answer these questions accurately and instantly.
Use AI to find and implement cost reduction suggestions
Foundational automatically identifies cost-saving opportunities in your data warehouse (see Cost Savings Recommendations). By connecting your AI tool via the MCP server, you can interactively explore these suggestions. Ask the AI: "Are there any expensive tables that are unused?", "Explain why this pipeline is flagged for running too frequently", or "Show me the lineage for this table recommended for deletion to confirm it has no downstream consumers".
Find potential security and privacy issues
Leverage AI to proactively identify data governance risks. Ask the AI "List all tables containing columns with pii tag and show their downstream consumers", "Are there any data pipelines joining columns tagged with pii tag with data tagged with public tag?". The MCP server provides the AI with the necessary lineage context from Foundational, enabling rapid identification of potential compliance issues.

Configuring Foundational MCP

Using Foundational MCP Server requires a user in Foundational App.

Claude Web

Follow the instructions in Anthropic site, under: "Adding Integrations to Claude".

Set https://mcp.foundational.io/mcp as the integration's remote MCP server URL.

Other Clients

It’s worth noting that most MCP clients currently do not support the latest MCP spec, so for our examples we might use a bridge client such as npx mcp-remote. We recommend you use it as well, and we’ll show our examples using it.

When you use npx mcp-remote, or a similar bridge client, you need to make sure it's running on the fixed port 8818:

{
  "mcpServers": {
    "foundational-io": {
       "command": "npx",
       "args": ["mcp-remote", "https://mcp.foundational.io/mcp", "8818"]
     }
  }
}

Install Node and `mcp-remote` package

In order to use remote MCP, you must have the both Node.js installed, as well as the mcp-remote package.

If you do not currently have Node.js installed, please visit the official Node.js website at https://nodejs.org/ and download the LTS (Long-Term Support) version suitable for your operating system.
Note: the version of Node must be above v18.

Once Node.js has been successfully installed, open your terminal or command prompt and execute the following command to install the mcp-remote package globally on your system:

npm install -g mcp-remote

Claude Desktop

Follow the instructions in Anthropic site, under: "Adding Integrations to Claude".

Cursor

Follow the instructions in Cursor documents, under: "Configuring MCP Servers".

VS Code

Follow the instructions in Visual Studio Code documents.

Troubleshooting

If the MCP server fails to load, you will usually see an error indication in the UI or in the logs:

In these cases, try to open the logs if you can. The logs path depend on the tool itself, for example, for Claude on Mac it would be under: /Users/<your-user-name>/Library/Logs/Claude/mcp-server-foundational-io.log

The following paragraphs talk about specific error messages and how to resolve those

TransformStream Error

The logs will usually show error like:

var EventSourceParserStream = class extends TransformStream {
                         ^ReferenceError: TransformStream is not defined

or, something on the node version:

npm warn cli npm v10.9.2 does not support Node.js v16.18.1. This version of npm supports the following node versions: `^18.17.0 || >=20.5.0`. You can find the latest version at https://nodejs.org/.

In these cases, the problem is usually that you have an installed Node version that is less than the required v18. You should remove the older Node version, by following these steps:

See all installed Node versions:
```
 ls -la ~/.nvm/versions/node/
```
If you see versions that are less than v18, you will need to uninstall those by using this command:
```
nvm uninstall v16.18.1
```

Using Foundational MCP With Your AI Tools