> For the complete documentation index, see [llms.txt](https://docs.subsalt.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.subsalt.io/retrieving-data/mcp.md).

# MCP

Subsalt exposes a **Model Context Protocol (MCP)** server so AI agents — Claude, Cursor, and any other MCP-compatible client — can work with your data through a single authenticated connection. An agent connected to Subsalt can explore what data exists, iterate on fully synthetic data, and compute high-precision answers on real data that are legally de-identified before they ever reach the model.

The guiding principle is the same one behind the rest of Subsalt: **neither the agent nor the user ever touches a sensitive row.** The MCP server only ever returns metadata, synthetic data, or privacy-banded aggregates.

{% hint style="info" %}
MCP is an open standard for connecting AI applications to external tools and data. If your client supports MCP, it can connect to Subsalt with no custom code — see connecting-a-client.md.
{% endhint %}

#### What an agent can do

The server groups its tools into three components, which together support the full analyst workflow:

<table><thead><tr><th width="220">Component</th><th>What it provides</th></tr></thead><tbody><tr><td><strong>Schema &#x26; metadata</strong></td><td>The canonical description of what data exists — databases, tables, and columns with types, descriptions, sample values, and foreign keys. Safe to read; it describes structure, not sensitive values.</td></tr><tr><td><strong>Synthetic data</strong></td><td>Connection details for every database's representative synthetic data over a standard PostgreSQL interface — a fast, safe loop for exploring distributions and developing an analysis.</td></tr><tr><td><strong>Secure Compute</strong></td><td>A way to run an analysis script against <em>real</em> data inside an isolated environment and get back only privacy-banded aggregates — a precise answer that is legally de-identified under HIPAA.</td></tr></tbody></table>

For the complete tool list and the recommended workflow that ties these together, see tools-and-workflow\.md.

#### How it fits with the rest of Subsalt

The MCP server is an alternative front door to the same platform described elsewhere in these docs. It authenticates with your normal Subsalt identity (including SSO), and every request it makes is authorized exactly as it would be through the portal or the PostgreSQL endpoint — an agent can only see the databases you can see.

* Prefer a SQL client or the portal for direct, interactive querying of synthetic data? See running-queries.md.
* Want an AI-assisted, natural-language path to a synthetic view without wiring up an agent? See data-request.md.

#### Getting started

1. Connect your MCP client to the Subsalt MCP endpoint and sign in.
2. Ask the agent to call the **`instructions`** tool first — it returns a getting-started guide and lists the databases available to you.
3. Explore, iterate on synthetic data, and confirm answers against real data with Secure Compute, following the standard workflow.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.subsalt.io/retrieving-data/mcp.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.