This project provides a framework for integrating 1C:Enterprise data and functionality with AI assistants by implementing the Model Context Protocol (MCP). It targets developers needing to leverage AI for tasks within their 1C systems, offering a pre-built 1C extension and an optional Python proxy to streamline context gathering and tool execution for LLMs.

How It Works

The core of the solution is a 1C:Enterprise configuration extension that acts as an MCP server. This extension exposes 1C data and business logic as "tools" that AI models can dynamically invoke. The Model Context Protocol (MCP) enables AI assistants to automatically request necessary context by calling these tools, rather than relying on manually prepared data. A Python proxy is recommended to handle infrastructure challenges like stdio transport and secure authentication via OAuth2.

Quick Start & Requirements

1. Install 1C Extension: Attach the pre-built MCP_Сервер.cfe extension to your 1C configuration.
2. Publish HTTP Service: Expose the mcp_APIBackend HTTP service from the extension on a web server.
3. Connect AI Client: Configure AI clients (e.g., Cursor) to connect to the published HTTP service endpoint.
   • Prerequisites: A running 1C:Enterprise platform, a web server capable of hosting HTTP services.
   • Security Note: Direct connection without authentication is insecure and limits client compatibility. The recommended approach involves the Python proxy.
   • Python Proxy Setup: Detailed instructions are available in the Python proxy documentation, including Docker deployment options.

Highlighted Details

• Provides "out-of-the-box" tools for working with 1C configuration metadata.
• Supports multiple connection variants: direct (less secure, limited) and via a recommended Python proxy (handles stdio transport, OAuth2 authentication).
• Extensible by developing custom 1C tools, resources (for static context like files), and prompts (for message templates).

Maintenance & Community

The project is actively developed, with contributions and suggestions welcomed via GitHub Issues.

Licensing & Compatibility

Licensed under the MIT License, generally permissive for commercial use and integration.

Limitations & Caveats

Directly connecting AI clients to the 1C HTTP service without authentication is insecure and not recommended. Some AI clients requiring stdio transport cannot be used in direct connection mode. The Python proxy is optional but highly recommended for robust and secure integration.