SkillFoxx

VK bot with a product MCP server

Product MCP + VK Бот

A learning project: a natural-language VK bot searches and edits products in SQLite through its own REST tool server and OpenAI

  • 0 ★ on the repository
  • Last commit
MCP server

Extended access

Why this level

  • The bot replies to VK community customers automatically without moderation
  • Product edit and delete tools are reachable through the LLM's interpretation of a request
All reasons and checks
Russian stack

gvoronin7-star/vk-bot-with-mcp-serverAuthor:gvoronin7-star

Install

Services

Manual install

Text for your agent

Clone the repository, install dependencies separately in mcp_server and VK_bot with pip install -r requirements.txt, create a .env with VK_API_TOKEN, PROXY_OPENAI_API_KEY, MCP_SERVER_URL and OPENAI_BASE_URL, and run mcp_server/server.py and VK_bot/bot.py in two terminals.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The bot replies to VK community customers automatically without moderation
  • Product edit and delete tools are reachable through the LLM's interpretation of a request

This is third-party code. Review the repository files before installing.

What it does

The project builds a demo shop: a SQLite database of 102 products across 6 categories, a FastAPI tool server with GET /tools and POST /call endpoints for search, add, edit, delete and a safe calculator with pagination, and a VK bot that receives messages via VK Long Poll, resolves intent through OpenAI (gpt-4o-mini via the proxyapi.ru proxy) and calls the matching server tool. The author labels the server side an MCP server, but it is the project's own REST protocol with tools and call endpoints, not a standard JSON-RPC MCP implementation, so it cannot be wired directly into Claude Code, Cursor or another MCP client without extra work.

Who it is for. For people studying LLM tool-calling bot architecture through a VK example and looking for a ready-made learning scaffold.

Good fit when

  • You want a learning example of a VK bot, an LLM and a tool server tied to a SQLite database
  • You want a scaffold you can rework for your own catalog and messenger
  • You are interested in the specific implementation: Long Poll, a 10-message dialog memory, a safe calculator

Not a fit when

  • You are looking for a standard MCP server to plug into Claude Code or Cursor: this protocol is custom, not JSON-RPC MCP
  • You need a production-ready shop rather than a learning project with 102 sample products

Example request

In a message to the bot: find a smartphone between 5000 and 10000 rubles in the electronics category

MCP

Transport
http
Authentication
not required

Environment variables

VK_API_TOKEN required, secret
A VK community token for receiving messages via Long Poll.
PROXY_OPENAI_API_KEY required, secret
An OpenAI key used through the proxyapi.ru proxy service.
MCP_SERVER_URL required
The address of the running tool server, defaults to http://127.0.0.1:8000.

Limitations

Despite the name, the tool server does not implement the MCP protocol: it is a custom REST API with GET /tools and POST /call, built for this project's own VK bot. It needs a VK community token and an OpenAI key routed through the proxyapi.ru proxy, whose availability is not guaranteed. The database is a sample one, and no license is stated in the repository.

How to disable. Stop the server.py and bot.py processes and delete the .env file with tokens.

FAQ

Can this MCP server be plugged into Claude Code as a normal MCP?

Not directly: the server uses its own /tools and /call endpoints rather than the JSON-RPC MCP protocol standard MCP clients expect.

Do I need a separate OpenAI key?

Yes, the project uses OpenAI gpt-4o-mini through the proxyapi.ru proxy service, with the key set in .env.

Discussion

No questions or reviews yet.

All VK tools
Russian stack

A VK API MCP server for business: community shop, posts, ads, stats, messages and lead forms with write confirmation

MCP serverBroad access3
Russian stack

MCP server auto-generated from the VK API schema: over 180 tools with filters by read, write, money modes and API sections

MCP serverBroad access1
Russian stack

MCP for VK on your own server: a VK API method catalog, reading walls and communities, community comparison and writes only after a human approves

MCP serverBroad access
Russian stack

A VK Ads MCP server with a turnkey installer: 136 tools, VK community research and a skill for several agents at once

MCP serverBroad access
Russian stack

MCP server for the VK API: profiles, walls, communities, photos, likes and stats, with confirmable post publishing

MCP serverBroad access17
Russian stack

Local MCP server for the VK API: search for people and cities, profiles, photos and video with an HTML gallery, a token pool and a SQLite cache

MCP serverBroad access

README in short

The README frames the project as solving manual catalog management and repetitive customer questions in a VK community: the bot answers instantly, and an admin edits the assortment without database access. It lists the stack (Python, FastAPI, SQLite, vk_api, OpenAI via proxyapi.ru), a four-block architecture, demo screenshots, setup instructions across two virtual environments, and a table of example bot commands.

README badge

Report a problem