vladimir-kharin/1c_mcp

What it solves

It enables AI assistants (such as Claude Desktop, Cursor, and other LLM clients) to interact directly with 1C:Enterprise databases. By implementing the Model Context Protocol (MCP), it allows AI models to automatically request specific data and execute functions within a 1C system rather than relying on manually provided context.

How it works

The project provides a 1C extension that acts as an MCP server, handling the technical protocol requirements. It exposes an HTTP service (mcp_APIBackend) that AI clients can call. To solve authentication and transport issues (like stdio support), the project includes an optional Python-based proxy that sits between the AI client and the 1C HTTP service, supporting OAuth2 and Basic Auth.

Who it’s for

Developers working with the 1C:Enterprise platform who want to integrate their business data and logic into AI-powered workflows or agents.

Highlights

• Ready-to-use Extension: Includes a pre-built 1C extension to handle MCP protocol overhead.
• Out-of-the-box Tools: Provides built-in tools for working with 1C configuration metadata.
• Flexible Deployment: Supports direct HTTP connection or a Python proxy (available via Docker) for better security and compatibility.
• Extensible Architecture: Allows developers to create custom tools by implementing simple export methods in 1C processing modules.
• Full MCP Support: Supports not just Tools, but also Resources (static context) and Prompts (message templates).