1C Export Engine Without the Platform Beats ibcmd by 5.3x, and an AI Agent Is Open-Sourced Under MIT
Infostart has rounded up five 1C developments. The most striking by the numbers is a custom engine that converts a CF configuration file into XML without the platform or a database in 4.3 seconds. That is 5.3 times faster than the standard ibcmd utility, and all 9,174 files of BSP 3.1 matched the reference output. Alongside it is an open-source AI agent for "1C:Accounting 3.0" under the MIT license, which explains why a document will not post. For 1C teams, this is a reason to try such utilities on a test database.
AI-processed from Habr AI; edited by Hamidun News
Picture a typical day for a 1C developer: you need to export the configuration to XML so you can put it in git and run automated checks, and that means spinning up the platform and an information database. One Infostart author built his own engine that converts a CF configuration file into XML without the platform and without a database. He exported the entire BSP 3.
1 (the standard 1C Subsystems Library) in 4.3 seconds. That is 5.
3 times faster than the standard ibcmd utility that ships with 1C. All 9,174 files matched the output of the standard export, with no errors. This is one of five publications in the latest roundup.
The others cover an AI agent for accounting under the MIT license (which allows anyone to freely use and modify the code), a review of MCP servers, database reconciliation, and the new WebSocket mechanism.
An AI Consultant and a Map of MCP Tools
The first publication, "AI Agent for Accounting," describes an open prototype of an agent for "1C:Accounting 3.0." It runs directly from the 1C interface. The agent analyzes debts and turnovers, checks the state of the books, explains why a document will not post, and creates documents on request. The author describes the architecture, how to connect local and cloud language models, the security mechanisms, and how the agent's actions are controlled. The code is open under the MIT license, so the project can serve as a foundation for your own work. The author is also candid about the limitations: it is a prototype, and there are scenarios it cannot yet handle.
The second piece, "AI Agents + 1C Databases via MCP: A Map of 2026 Tools," compares 13 solutions that appeared on Infostart in 2026. MCP (Model Context Protocol) is a way to give a neural network access to an external system through a standard interface. At first, such servers could only read metadata and run queries. Now some of them edit code, load changes into the configuration, and check the result in the information database. The authors compare the solutions by maturity, connection methods, and impact on the configuration, and separately examine GitHub projects and security issues. If you are choosing a server for an agent, this map can save you several evenings of experimenting.
Database Reconciliation and WebSocket
The third publication, "Comparing and Reconciling Data Between 1C Databases," offers a universal external data processor. It compares two databases by catalogs, documents, registers, attributes, and tabular sections. Fields are matched automatically, link keys are configured manually, results can be filtered, and comparison settings are saved in XML. It works locally and over a COM connection. Differences are shown as a tree, down to specific values. It is useful for migration, for setting up data exchanges, and for verifying that data was transferred without loss.
The fourth, "WebSockets Don't Hurt," covers a mechanism that appeared in the 1C 8.3.27 platform. WebSocket lets a client and server exchange data in both directions without endless HTTP requests along the lines of "anything new?". The author shows three connection options, authorization setup, event handling, and testing. He also warns about platform bugs and limitations of the new mechanism. Use cases include chats, notifications, hardware integration, database synchronization, and real-time data updates.
What to Take Away in Practice
One idea ties the roundup together: a layer of tools that do not wait for official platform updates is growing fast around 1C. Here is a practical order of attack. If you want AI in your accounting system, start with the review of MCP servers and study the open-source agent, paying attention to its mechanisms for controlling actions.
If you handle migrations, take a look at the database reconciliation processor. If you build integrations, test WebSocket on 8.3.
27 and account in advance for the limitations described. And if you run heavy CI around configurations, keep an eye on the export engine: the 5.3x speedup was measured in the author's own test, so results may differ on your hardware and configurations.
A caveat: these are announcements of publications, not independent tests. The figures and capabilities come from the authors' descriptions, and the maturity of each solution should be checked on a test database before it touches production data, especially if the agent can create documents.
- Nov 19, 2026Open DaisyRose-27B solves 12% of Terminal-Bench tasks, up from 5.6%
- Nov 19, 2026The same Gemma 4 wrote a backend that passed 80% of checks in opencode, but only 41% in OpenHands
- Nov 19, 2026OpenAI's coding agents would cost $2400 a day per researcher at API prices
- Nov 19, 2026AWS AgentCore Agents Can Now Reach a Private Keycloak Behind a Corporate CA
Key Questions Answered
What are the key details on An AI Consultant and a Map of MCP Tools?
The first publication, "AI Agent for Accounting," describes an open prototype of an agent for "1C:Accounting 3.0." It runs directly from the 1C interface. The agent analyzes debts and turnovers, checks the state of the books, explains why a document will not post, and creates documents on request. The author describes the architecture, how to connect local and cloud language models, the security mechanisms, and how the agent's actions are controlled.
What are the key details on Database Reconciliation and WebSocket?
The third publication, "Comparing and Reconciling Data Between 1C Databases," offers a universal external data processor. It compares two databases by catalogs, documents, registers, attributes, and tabular sections. Fields are matched automatically, link keys are configured manually, results can be filtered, and comparison settings are saved in XML. It works locally and over a COM connection. Differences are shown as a tree, down to specific values.
What to Take Away in Practice?
One idea ties the roundup together: a layer of tools that do not wait for official platform updates is growing fast around 1C. Here is a practical order of attack. If you want AI in your accounting system, start with the review of MCP servers and study the open-source agent, paying attention to its mechanisms for controlling actions. If you handle migrations, take a look at the database reconciliation processor.
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.