> ## Documentation Index
> Fetch the complete documentation index at: https://documentation.idenfy.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Context Protocol (MCP)

> Install and configure the iDenfy Model Context Protocol server to give AI assistants like Cursor, Claude, and Windsurf access to these docs.

This guide explains what the iDenfy MCP server is, how to install it, and how to use it when you work with iDenfy.

## What is MCP

[Model Context Protocol](https://modelcontextprotocol.io) is an open standard that lets AI assistants connect to external tools and data sources. The iDenfy MCP server gives your AI assistant access to this documentation. Once it is connected, the assistant can search the docs and read full pages while it answers your questions or writes code.

<Note>
  The iDenfy MCP server is read-only. It only serves public documentation. It does not have access to your API keys, dashboard, or verification data.
</Note>

## Works with Every iDenfy Product

The MCP server covers the full iDenfy documentation, so you can use it for any product in our platform:

* [Identity Verification (KYC)](/kyc/overview)
* [Business Verification (KYB)](/kyb/overview)
* [AML Screening and Monitoring](/aml/overview)
* [Fraud Prevention](/fraud-prevention/overview)
* [Face Authentication](/face-authentication/overview)
* [Bank Verification](/bank-verification/overview)
* [Mobile SDKs](/sdks/overview), [API Reference](/api-reference/overview), and every guide under [Dashboard](/guides/dashboard/overview) and [Compliance](/guides/compliance/overview)

You can ask your AI assistant a question about any of these areas and it will search the right pages for you.

## What You Can Do with It

<CardGroup cols={2}>
  <Card title="Search the docs" icon="magnifying-glass">
    Ask questions in plain language. The assistant searches every guide and API reference for you.
  </Card>

  <Card title="Read full pages" icon="file-lines">
    Get the full content of a page by its path when a short snippet is not enough.
  </Card>

  <Card title="Write integration code" icon="code">
    Ask the assistant to draft KYC, KYB, or webhook code based on the current docs.
  </Card>

  <Card title="Debug your setup" icon="bug">
    Look up error codes, webhook fields, or parameter names without leaving your editor.
  </Card>
</CardGroup>

## Available Tools

The iDenfy MCP server has two tools:

| Tool                            | What it does                                                                      |
| ------------------------------- | --------------------------------------------------------------------------------- |
| `search_idenfy_documentation`   | Searches the docs and returns page titles, paths, and short snippets.             |
| `get_page_idenfy_documentation` | Returns the full content of a page by its path, for example `kyc/generate-token`. |

The usual flow is simple: **search first** to find the right page, then **read the full page** if you need more detail.

## Install the MCP Server

You do not need to install anything on your computer. The iDenfy MCP server runs online. You only need to add the server URL to your AI assistant.

<Info>
  Server URL: `https://documentation.idenfy.com/mcp`
</Info>

### Claude Desktop

Open the [Claude Desktop](https://claude.ai/download) configuration file `claude_desktop_config.json` and add the server:

```json theme={"system"}
{
  "mcpServers": {
    "idenfy-docs": {
      "url": "https://documentation.idenfy.com/mcp"
    }
  }
}
```

Restart Claude Desktop. The iDenfy tools will show up in the tool list.

### Claude Code

Run this command in your [Claude Code](https://claude.com/claude-code) terminal:

```bash theme={"system"}
claude mcp add --transport http idenfy-docs https://documentation.idenfy.com/mcp
```

If you want to share the setup with your team, add it to `.mcp.json` in your project root:

```json theme={"system"}
{
  "mcpServers": {
    "idenfy-docs": {
      "url": "https://documentation.idenfy.com/mcp"
    }
  }
}
```

### Cursor

In [Cursor](https://cursor.com), open **Settings > MCP > Add new MCP server** and paste the server URL. The server will be ready to use in every workspace.

### ChatGPT

Custom connectors are available on [ChatGPT](https://chatgpt.com) Pro, Business, Enterprise, and Edu plans.

1. Open **Settings > Connectors** in ChatGPT.
2. Click **Add custom connector**.
3. Set the name to `iDenfy Docs` and the MCP server URL to `https://documentation.idenfy.com/mcp`. Leave authentication set to **None**.
4. Save the connector.
5. In a new chat, open the tools menu and turn on the **iDenfy Docs** connector.

<Note>
  On Business or Enterprise workspaces, a workspace admin may need to allow custom connectors before you can add one.
</Note>

### Google Gemini

You can connect the iDenfy MCP server to [Gemini](https://gemini.google.com) through the Gemini CLI or Gemini Code Assist.

Open your Gemini settings file (for example `~/.gemini/settings.json`) and add the server:

```json theme={"system"}
{
  "mcpServers": {
    "idenfy-docs": {
      "url": "https://documentation.idenfy.com/mcp"
    }
  }
}
```

Restart Gemini. The iDenfy tools will show up in the tool list.

### Perplexity

[Perplexity](https://www.perplexity.ai) supports MCP servers through **Connectors** in the Perplexity desktop app and the Comet browser.

1. Open **Settings > Connectors** in Perplexity.
2. Click **Add connector** and select **Custom MCP server**.
3. Set the name to `iDenfy Docs` and the URL to `https://documentation.idenfy.com/mcp`.
4. Save the connector and turn it on for the spaces where you want to use it.

### Other Clients

Any tool that supports MCP can connect to the iDenfy server with the same URL. This includes [Zed](https://zed.dev), [Windsurf](https://windsurf.com), the [VS Code](https://code.visualstudio.com) MCP extension, and custom agents. Check your tool's documentation for the exact steps.

## How to Use It

After you connect the server, you just talk to your assistant in plain language. You do not need to call the tools yourself. The assistant decides when to search the docs or read a full page.

### Find Information in the Docs

```text theme={"system"}
How does iDenfy liveness detection work, and which parameters control it?
```

The assistant searches the docs, reads the right page, and tells you about active and passive liveness with the correct parameter names.

### Generate Integration Code

```text theme={"system"}
Write a Python script that creates a KYC session with name and date of birth
matching, and checks the webhook signature.
```

The assistant uses the [Create Verification Session](/kyc/generate-token) and [Callback Signing](/security/callback-signing) pages, so the code uses the right endpoint, the right fields, and the correct HMAC check.

### Debug a Failing Webhook

```text theme={"system"}
My webhook handler returns 403 for every iDenfy request. What are the common
reasons for a signature mismatch?
```

The assistant opens the troubleshooting section of the callback signing page and explains raw body handling, hex encoding, and proxy issues.

### Answer Compliance Questions

```text theme={"system"}
Which documents does iDenfy support for Lithuanian citizens, and what are the
NFC requirements for eIDAS high assurance?
```

The assistant searches the supported documents reference and the eIDAS guide and gives you one answer with the right sources.

## Best Practices

<AccordionGroup>
  <Accordion title="Be specific in your prompts">
    The more detail you give, the better the search result. Instead of *"How do I create a session?"*, ask *"How do I create a KYC session with restricted countries and a custom callback URL?"*.
  </Accordion>

  <Accordion title="Ask for the page it used">
    Ask the assistant to include the page path or link in the answer. This way you can open the source and check the answer yourself.
  </Accordion>

  <Accordion title="Use it together with your code">
    In an IDE like Claude Code or Cursor, let the assistant read both the iDenfy docs and your project. The code it writes will match your existing style instead of being generic.
  </Accordion>

  <Accordion title="Do not share API keys with the assistant">
    The iDenfy MCP server does not return credentials or customer data. Keep it that way on your side: do not paste your API key or secret into the chat. Use environment variables in the code instead.
  </Accordion>

  <Accordion title="Search before you read full pages">
    A full page uses more of the assistant's memory than a search snippet. Let the assistant search first, and only read the full page if the snippet is not enough.
  </Accordion>

  <Accordion title="Always review the generated code">
    MCP reduces guesses because the assistant reads the real docs, but you should still review the code. Check signature verification, raw body handling, and how credentials are stored before you deploy.
  </Accordion>
</AccordionGroup>

## Example: Build a KYC Integration from Scratch

You can build a full KYC integration in one conversation with your assistant:

<Steps>
  <Step title="Ask for the flow">
    *"Explain the iDenfy KYC flow from session creation to webhook result."* The assistant describes the three steps: [create a session](/kyc/generate-token), send the user to the UI or SDK, and handle the [webhook result](/kyc/webhooks).
  </Step>

  <Step title="Create the session">
    *"Write the Node.js code that creates a session with data matching and returns `redirectUrl`."* The code uses the current endpoint and field names.
  </Step>

  <Step title="Handle the webhook">
    *"Add an Express handler that checks the `Idenfy-Signature` header against the raw body."* The code follows the pattern from the [callback signing](/security/callback-signing) page.
  </Step>

  <Step title="Clean up the parameters">
    *"Which of these parameters are already set in the dashboard by default?"* The assistant checks the [Create Verification Session](/kyc/generate-token) parameter list and removes the ones you do not need.
  </Step>
</Steps>

The result is a working draft based on the current docs, not old training data.

## Privacy and Scope

* The MCP server only serves content from [documentation.idenfy.com](https://documentation.idenfy.com).
* It does not have access to your dashboard, sessions, API keys, or customer data.
* Your prompts and the doc content go through your AI assistant provider (for example [Anthropic](https://www.anthropic.com), [OpenAI](https://openai.com), [Google](https://ai.google), or [Perplexity](https://www.perplexity.ai)). Check your provider's data policy before you share sensitive information.

## Feedback

If a search returns old or wrong results, or if a page is missing content you need, contact [support@idenfy.com](mailto:support@idenfy.com). Share the prompt you used and the page you expected to find.