# gigachat-mcp

**Category:** 🗣️ Conversational AI  
**Repository:** https://github.com/theYahia/gigachat-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 243  
**Views:** 9  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/gigachat-mcp

## Description
MCP server for Sber GigaChat API — chat completions, models, embeddings.

## Tools
Capabilities this server exposes over MCP:

- **chat** — Генерация текста через GigaChat. Поддерживает диалог с системным промптом.
- **list_models** — Список доступных моделей GigaChat.
- **embed_text** — Получение векторных представлений (эмбеддингов) текстов через GigaChat.
- **get_token_count** — Подсчёт токенов для массива текстов.
- **generate_image** — Генерация изображения через GigaChat (описание в тексте сообщения).
- **get_balance** — Проверка остатка токенов/квоты GigaChat.
- **list_assistants** — Список доступных ассистентов GigaChat.
- **file_upload** — Загрузка файла в GigaChat для мультимодальной обработки.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "gigachat-mcp": {
    "command": "npx",
    "args": ["-y","@theyahia/gigachat-mcp"],
    "env": {
      "GIGACHAT_AUTH_KEY": ""
    }
  }
}
```

**Requires environment variables:** `GIGACHAT_AUTH_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What gigachat-mcp MCP server does

The gigachat-mcp MCP server exposes Sber GigaChat API functions as MCP tools. It is intended for MCP-compatible clients that need to send conversational requests, inspect available models, process text, generate images, check account usage, work with assistants, or upload files for multimodal processing.

The package provides eight tools:

- `chat` generates text using a conversational exchange.
- `list_models` returns the available GigaChat models.
- `get_token_count` counts tokens in text.
- `generate_image` requests an image from GigaChat.
- `get_balance` reports remaining tokens or quota.
- `list_assistants` lists available assistants.
- `file_upload` uploads a file for multimodal processing.

## How it works

The server runs as a local process started through `npx` and communicates with the MCP client using the configured server command. The client supplies the `GIGACHAT_AUTH_KEY` environment variable, and the server uses it to authenticate with Sber GigaChat API. The key must be a Base64-encoded `client_id:client_secret` value obtained from Sber’s developer portal.

OAuth tokens are refreshed automatically. Once configured, an agent can invoke the exposed tools through its MCP interface rather than calling the GigaChat API directly.

## Setup and configuration

Install the package on demand with the npm package name `@theyahia/gigachat-mcp`; the documented configurations use `npx -y` and do not require a separate global installation.

For Claude Desktop, add an MCP server entry with `command` set to `npx`, arguments `-y` and `@theyahia/gigachat-mcp`, and `GIGACHAT_AUTH_KEY` in the server’s environment. Claude Code uses the equivalent command-line registration:

```bash
claude mcp add gigachat -e GIGACHAT_AUTH_KEY=your-auth-key -- npx -y @theyahia/gigachat-mcp
```

VS Code and Cursor use a server configuration with the same command, arguments, and environment variable. Replace the placeholder with the authentication key from developers.sber.ru.

## Tools and capabilities

Use `chat` for conversational text generation and `embed_text` when an application needs vector representations of text. `get_token_count` can help inspect text size before sending requests. `generate_image` handles image requests through GigaChat.

Operational and account-oriented tasks are also covered. `list_models` shows model choices, `get_balance` checks remaining tokens or quota, and `list_assistants` retrieves assistants. `file_upload` makes a file available for multimodal processing.

## Limitations and notes

The original repository is archived, and further development has moved to the `theYahia/WWmcp` monorepo. The current package source is identified there under `servers/gigachat/`. Issues and pull requests should be directed to that repository rather than the archived one.

The npm package remains `@theyahia/gigachat-mcp` and is described as unchanged.

_Full upstream README: https://allmcps.com/mcp/gigachat-mcp/readme_

