I'm looking for a person (not a bot) who understands AI agents and knows how to build them. By AI agent, I mean: processing input data, making a request to a 1x LLM or similar AI model, potentially requesting MCP or similar, potentially requesting a RAG system, processing output data And experience should include working with multiple agents that also self-check, pass data to each other, etc. Understanding of RAG systems is also required. Experience in automating marketing projects is preferable, for example: search for the latest topics using deep research on sources 1,2,3.... compare them with what has been found before and remove duplicates. Also compare with already published articles analyze the found news in relation to the necessary context search for and assess the importance of these topics create a report for the user create a sample post for LinkedIn, Instagram, X in the style that was previously used (and check)
Task Description: Development of MCP Server for 1C EcosystemGeneral Goal To develop an intermediary layer (MCP Server) that will allow LLM agents to safely interact with the 1C information base. This will enable users to receive reports, create documents, and analyze data through a chat interface using natural language.Functional Capabilities (Tools) The server must provide AI models with a set of tools for performing the following actions: Data Reading: Searching for counterparties, obtaining stock balances, sampling product prices. Analytics: Generating management reports in text or tabular form. Actions: Creating drafts of documents (Customer Order, Invoice), changing task statuses. Metadata: Obtaining the structure of objects (what fields are in the "Employees" directory) so that the model understands what it is working with.Main Implementation Stages API Design in 1C: Preparation of HTTP services in the 1C extension that will accept requests from the MCP server. Setting up authorization (Basic or Bearer token). Development of MCP Server: Defining the input parameter schema for tools (JSON Schema). Mapping requests from MCP to OData calls or HTTP requests to 1C. Security and Limitations: Restricting access rights (Read-only by default). Limits on the volume of returned data (to avoid "crashing" the model's context window with a huge table). Testing: Connecting the server to Claude Desktop or another MCP client. Checking scenarios: "How much of the product 'Brick' is left in the main warehouse?" or "Create a draft invoice for LLC 'Vector' for 5 monitors".Expected Result A working executable file or service that registers in the MCP client configuration. When a request is entered in the chat, the AI automatically calls the necessary tool, queries 1C, and provides a structured response based on real data from the system. Important Note: The main value of this solution lies in the transition from manual report generation to the concept of "Talk to your ERP" (talking to your ERP system).