Semantic search over 1C platform help
1c-sntx-sem
An MCP server for Cursor and Claude Desktop with hybrid embeddings and BM25 search over 1C platform help, the query language and the standard subsystems library
Low risk
We rate an entry low when it mostly gives the agent instructions and reference material.
Why this level
- Only reads platform help and BSP XML exports, does not connect to a production infobase
- May send query text to an external embeddings API if a non-local provider is chosen
Install
Manual install
pip install -e ".[embeddings]"Installs with the local E5 embedding model; without the extras only the external OpenAI-compatible API option works.
This is third-party code. Review the repository files before installing.
What it does
The server builds a search index over 1C platform help (BSL and the SDBL query language) and over exported methods of standard subsystems library (BSP) common modules exported to XML. Search is hybrid: embeddings plus BM25 combined with RRF; embeddings default to a local E5 model, with options via an OpenAI-compatible API and Ollama. Each user builds the help database locally from their own licensed platform with the ingest command; the database is not part of the repository. There is an HTTP API with a browser search UI and a Docker option where the API runs in a container while a thin MCP client runs on the host.
Who it is for. For 1C developers on Cursor or Claude Desktop who need meaning-based, not just text, search over built-in platform help and the standard subsystems library.
Good fit when
- You need meaning-based search over 1C platform help, not exact term matching
- You work with the standard subsystems library and need examples of exported methods
- You are ready to build the index locally from your own licensed platform
Not a fit when
- No licensed 1C platform is installed to build the index from
- 8 GB-plus RAM for the Docker IVF index is not available, and accuracy is lower without it
Example request
Look up how a left join works in the 1C query language and show an example from the standard subsystems libraryLimitations
The help database is not shipped with the repository; each user builds it from their own platform, which takes 5 to 15 minutes on CPU. The local E5 embedding model is about 120 MB and downloads on first run. Rebuilding the IVF index in Docker needs 8 GB or more of Docker Desktop RAM; without it, search runs without ANN.
How to disable. Remove the 1c-syntax-sem entry from the client's mcp.json and stop the containers with docker compose down if Docker was used.
MCP
- Transport
- stdio, http
- Authentication
- not required
| Environment variables | |
|---|---|
| SNTX_SEM_CONFIG | Path to config.yaml for the local in-process run. |
| SNTX_SEM_API_URL | API address for a thin MCP client on top of the Docker container. |
Security check
- Only reads platform help and BSP XML exports, does not connect to a production infobase
- May send query text to an external embeddings API if a non-local provider is chosen
README in short
The README gives a quick start: install with the embeddings extras, copy the config, the ingest command to build the index from the platform, ingest-bsp for the standard subsystems library, and scan-examples for examples from local configurations. It covers the HTTP API and web UI, a Docker mode splitting API and MCP, choosing an embedding provider among sentence_transformers, an OpenAI-compatible API or Ollama, plus architecture docs and a decision log.
FAQ
Is torch needed on the host when using Docker?
No, the Docker option has mcp.json.docker.example with no torch; the heavy computation stays in the API container.
Is the standard subsystems library indexed?
Yes, as a separate bsp domain: exported methods from the ПрограммныйИнтерфейс areas in common modules exported to XML.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything