---
metadata:
  - name: generator
    content: Diplodoc Platform v5.64.0
alternate:
  - https://yandex.com.tr/support/wiki/en/mcp.md
  - href: https://yandex.com.tr/support/wiki/en/mcp.md
    type: text/markdown
    title: Markdown version
  - href: https://yandex.com.tr/support/wiki/en/llms.txt?revision=r21825371
    rel: describedby
title: Yandex Wiki MCP server
description: How to connect Yandex Wiki to AI assistants and agents via an MCP server.
---
> **Documentation Index:** Fetch the complete configuration index at https://yandex.com.tr/support/wiki/en/llms.txt?revision=r21825371



# Yandex Wiki MCP server

[MCP (Model Context Protocol)](https://modelcontextprotocol.io/) is a standard protocol for connecting external tools to AI assistants and agents. Using the Wiki MCP server, you can give an AI assistant access to your organization's knowledge base.

## MCP server capabilities {#capabilities}

Via the MCP server, the assistant gets access to the [Wiki public API](https://yandex.com.tr/support/wiki/en/api-ref/about.md) and can:

- Search and read knowledge base pages
- Create and edit pages
- Collect summaries across multiple pages

## How to connect {#setup}

To connect the Wiki MCP server to an AI assistant:

1. Get an OAuth token for accessing the Wiki API. For more information, see [API access](https://yandex.com.tr/support/wiki/en/api-ref/access.md).

1. Find your organization ID. To do this, in [Yandex Tracker](https://tracker.yandex.com/admin/orgs), open the **Administration** → **Organizations** page and copy the **ID** field value.

1. Add the MCP server configuration to your AI tool's settings. For the parameters, specify the URL and headers from the table below.

Connection parameters:

#|
|| Parameter | Value ||
|| URL | `https://mcp.wiki.yandex.net` ||
|| Organization ID  |
One of the values:
* `X-Org-Id` header for an organization in Yandex 360 for Business
* `X-Cloud-Org-Id` header for an organization in Yandex Cloud ||
|| `Authorization` header |
One of the values:
* OAuth token in the `OAuth <token>` format
* IAM token in the `Bearer <token>` format — only for an organization in Yandex Cloud ||
|#

{% note info "" %}

For more information about authorization methods, see [Access to the API](https://yandex.com.tr/support/wiki/en/api-ref/access.md#headers).

{% endnote %}

## Connection examples {#connection-examples}

The method for adding an MCP configuration depends on the specific tool. Usually, it's enough to specify the MCP server URL and the required headers.

### Codex setup example {#codex-example}

To add the Wiki MCP server to Codex, add the following to the `.codex/config.toml` file:

```toml
[mcp_servers.wiki-oauth]
url = "https://mcp.wiki.yandex.net"
http_headers = {
    "X-Org-Id" = "<organization_ID>",
}
env_http_headers = {
    "Authorization" = "WIKI_OAUTH_TOKEN",
}
```

The `WIKI_OAUTH_TOKEN` environment variable must contain an OAuth token in the `OAuth <token>` format.

### Claude Code setup example {#claude-code-example}

To add the Wiki MCP server to Claude Code, run the following commands:

{% list tabs %}

- Unix

   ```bash
   export WIKI_OAUTH_TOKEN="OAuth <token>"
   ```

   ```bash
   claude mcp add --transport http wiki-oauth https://mcp.wiki.yandex.net \
       --header "X-Org-Id: <organization_ID>" \
       --header "Authorization: $WIKI_OAUTH_TOKEN"
   ```

- Windows

   ```cmd
   set "WIKI_OAUTH_TOKEN=OAuth <token>"
   ```

   ```cmd
   claude mcp add --transport http wiki-oauth https://mcp.wiki.yandex.net ^
       --header "X-Org-Id: <organization_ID>" ^
       --header "Authorization: %WIKI_OAUTH_TOKEN%"
   ```

{% endlist %}

The `WIKI_OAUTH_TOKEN` environment variable must contain an OAuth token in the `OAuth <token>` format.

### Yandex AI Studio setup example {#ai-studio-example}

To connect the Wiki MCP server through MCP Hub:

1. In [Yandex AI Studio](https://aistudio.yandex.ru/), select a folder.

1. In the left panel, expand **Agent Atelier** and select **MCP servers**.

1. Click **Create MCP server** and select **External MCP server**.

1. Enter a server name and click **Create**.

1. In the **URL** field, enter `https://mcp.wiki.yandex.net`, then select the `Streamable HTTP` transport.

1. In the **Authorization type** field, select **Access token** and add the headers:

   - for an organization in Yandex 360 for Business, `Authorization: OAuth <OAuth token>` and `X-Org-Id: <organization_ID>`;
   - for an organization in Yandex Cloud, `Authorization: OAuth <OAuth token>` or `Authorization: Bearer <IAM token>`, and `X-Cloud-Org-Id: <organization_ID>`.

1. Click **Connect**. In the **Tools** section, select the tools you need and click **Save**.

1. In the left panel, select **Agent Atelier** → **Agents**, then create or open an agent.

1. In the **Tools** field, click **Add**, select **MCP server**, and select the saved Wiki MCP server.

Check the connection with a read request:

> Find the `<page_title>` page, read it, and prepare a brief summary.

For more information, see the AI Studio instructions [on connecting an external MCP server](https://aistudio.yandex.ru/docs/ru/ai-studio/operations/mcp-servers/connect-external) and [creating an agent](https://aistudio.yandex.ru/docs/ru/ai-studio/operations/agents/create-agent-ui).

{% note info "" %}

The Wiki MCP server processes requests on behalf of the token user and with their permissions.

For an OAuth token, select the minimum permissions: `wiki:read` for reading pages, and `wiki:write` only for creating, updating, and deleting content.

If the Wiki administrator disables API access, the integration will not work. An IAM token is valid for no more than 12 hours.

{% endnote %}

{% note warning "" %}

You cannot use an AI Studio service account to authorize in Wiki.

Do not enable **Make MCP server public**, do not publish a gateway with a user token, and do not add real tokens to examples.

For tools that create, update, or delete pages, enable call confirmation. For more information about setting up the policy, see the [AI Studio documentation](https://aistudio.yandex.ru/docs/ru/ai-studio/concepts/agents/tools/mcp-tool).

{% endnote %}


<!-- source: en/_assets/style/image.md -->

<!-- endsource: en/_assets/style/image.md -->
