How to Connect 1C:Enterprise to ChatGPT via MCP

💡 No install? Use cloud.anythingmcp.com directly. Sign in, click Connectors → 1C:Enterprise, paste your credentials, mint an MCP API key and you're done. No Docker, no git clone, no local server.

1C:Enterprise + ChatGPT

Read your 1C:Enterprise 8.3 infobase (1С:Бухгалтерия, Управление торговлей, 1С:ERP, УНФ and other configurations) through its standard OData interface: catalogs, documents with their tabular sections, register balances, turnovers and slices, with the names from your configuration's own metadata. Read-only, Basic auth, your own server.

Prerequisites

See the full setup instructions baked into the connector (visible in the in-app store when you select the connector). The required environment variables for this connector are:

ONEC_BASE_URL, ONEC_USER, ONEC_PASSWORD

Step 1 — Get credentials

Setting up takes about 30 minutes and needs your 1C administrator. You need a 1C:Enterprise 8.3 infobase (platform 8.3.5 or newer) published on a web server with the standard OData interface switched on, and a dedicated 1C user for the connector.

1. Publish the infobase on a web server (Apache or IIS with the platform's web server extension modules). In the Designer open Администрирование → Публикация на веб-сервере (Administration → Publish to web server), give the publication a name in Latin letters, tick Публиковать стандартный интерфейс OData (Publish standard OData interface) and publish. If a 1C hosting provider runs your infobase (1cFresh, cloud rental), ask the provider to enable OData for it.
2. Choose what OData may show. Out of the box the interface exposes no objects. In typical configurations open Администрирование → Синхронизация данных → Настройки стандартного интерфейса OData (or Все функции → Обработки → Настройка стандартного интерфейса OData) and add the catalogs, documents and registers the AI may read. In other configurations a developer calls УстановитьСоставСтандартногоИнтерфейсаOData().
3. Create a dedicated 1C user with read-only roles for those objects and the right to use the standard OData interface. The settings form from step 2 can create this user for you.
4. Check in a browser: open your publication address followed by /odata/standard.odata/?$format=json and sign in with that user. A JSON list of objects means OData works; an empty list means step 2 is missing.
5. Paste the publication address into ONEC_BASE_URL: the address you open the web client with, without /odata and without a trailing slash, for example https://1c.example.com/trade. Put the user name into ONEC_USER exactly as in the infobase user list (Cyrillic names work) and its password into ONEC_PASSWORD.
6. Install the connector and run onec_enterprise_describe_service: it lists every object published to OData, with its keys. An empty list means nothing has been added to the OData composition yet.

Your server must be reachable: AnythingMCP Cloud only reaches addresses on the public internet, so publish over HTTPS on a public host; the password travels with HTTP Basic authentication. For a 1C server inside your network, self-host AnythingMCP there (or connect it through a VPN) and add the host to SSRF_ALLOWED_HOSTS.

What it reads: catalogs, documents with their tabular sections, information, accumulation and accounting registers, charts of accounts, constants and document journals, under the names from your configuration's own metadata, usually Russian: Catalog_Номенклатура, Document_РеализацияТоваровУслуг with its lines in Document_РеализацияТоваровУслуг_Товары, AccumulationRegister_ТоварыНаСкладах, AccountingRegister_Хозрасчетный. It works with every configuration on the platform: 1С:Бухгалтерия предприятия, Управление торговлей, 1С:ERP, Комплексная автоматизация, УНФ, Розница, ЗУП and custom ones.

Balances, turnovers and slices: for registers the AI reads the virtual tables Balance, Turnovers, BalanceAndTurnovers, SliceLast and SliceFirst with onec_enterprise_get, so 1C filters and aggregates on the server instead of the AI adding up thousands of document lines. That is how it answers stock at a date, sales in a period, the latest exchange rate or the balance of account 62.01.

How the AI works: onec_enterprise_guide explains names, filters and registers at no cost. onec_enterprise_describe_entity shows the fields of an object and their types, onec_enterprise_query reads lists with filter, sorting, selected fields and paging, and onec_enterprise_get_entity reads one object by its Ref_Key together with its tabular sections. Reference fields such as Контрагент_Key hold a GUID; the AI looks up the name instead of showing you GUIDs.

Clean totals: the AI counts only posted documents (Posted eq true) and leaves out objects marked for deletion unless you ask for them. Dates carry no time zone: they are the infobase's local time. Amounts are in the document currency unless the field says otherwise.

Read-only: no tool writes. The connector never creates, changes, posts or deletes anything, so there is nothing to confirm and no tool is switched off at install. The rights of the 1C user are the real boundary: 1C checks them on every call, and what the user may not see does not appear in the answer, which can look like missing data.

Limits: there is no fixed call limit, but every query runs on your 1C server. The AI always sets $top and $select on large catalogs and registers, and pages through long lists.

Common errors: 401 means 1C refused the sign-in: check the user name, the password and the right to use the standard OData interface. 403 means the user signed in but may not read that object: give its role read rights. 404 means the object is not in the OData composition, its name is misspelled, or ONEC_BASE_URL is wrong (it must not contain /odata). 400 with "Ошибка при разборе опции запроса" points to a wrong filter or field name. 500 on a file infobase usually means the web server account has no write access to the infobase folder.

The connector does not create, post or delete documents, and it only sees the objects your administrator added to the OData composition.

Step 2 — Install the adapter

curl -fsSL https://raw.githubusercontent.com/HelpCode-ai/anythingmcp/main/docker-compose.quickstart.yml -o docker-compose.yml
printf 'JWT_SECRET=%s\nENCRYPTION_KEY=%s\n' "$(openssl rand -hex 32)" "$(openssl rand -hex 32)" > .env
docker compose up -d

Step 3 — Add the connector in ChatGPT

Your server URL: in AnythingMCP, open MCP Servers → the server this connector is on, and copy its URL (https://cloud.anythingmcp.com/mcp/…). Use it wherever this guide shows YOUR_SERVER_ID.

ChatGPT adds MCP servers as plugins on its Plugins page, on the web and on paid plans. Developer mode is no longer needed. In a Business workspace only admins and owners can add one; in Enterprise and Edu an admin can grant the Create plugins with MCPs permission.

1. Go to chatgpt.com/plugins, click Add, then Add custom MCP server.
2. Enter a name, e.g. Name: 1C:Enterprise.
3. Under Connection, enter the Server URL https://cloud.anythingmcp.com/mcp/YOUR_SERVER_ID and keep Authentication on OAuth.
4. Select I understand and want to continue, then Create as a plugin. When ChatGPT asks, sign in to AnythingMCP and allow access.
5. Install the new plugin from your personal plugins, then type @ in a chat and pick it.

Available tools

Example prompts

• "What is the stock of each item at the main warehouse today, and which items are below 10 units?"
• "How much did we ship to Romashka LLC in September, broken down by posted sales document?"
• "Find the counterparty with INN 7707083893 and list the invoices we issued to it last month."
• "Which US dollar exchange rate is stored in 1C for today?"
• "What is the balance of account 62.01 at the end of September, and which customers make it up?"

FAQ

Does the free ChatGPT plan work with custom MCP servers? OpenAI documents custom MCP servers for paid plans only; the Free plan is not on its list.

Next steps

• Connector landing page
• Connect to Claude
• Connect to Cursor
• Connect to Gemini
• Connect to Meta Muse
• Connect to Microsoft Copilot