1c-templates-mcp
An MCP server with hybrid search over 2262 community BSL code templates and a CRUD web interface built on Monaco Editor
Low risk
We rate an entry low when it mostly gives the agent instructions and reference material.
Why this level
- Only works with the local template database, does not connect to a live 1C base
- The MCP endpoint and web interface are open with no authorization on the chosen port
Install
Manual install
cd deploy && docker compose up -dInstall from the ready desko77/1c-templates-mcp Docker Hub image, the recommended method.
This is third-party code. Review the repository files before installing.
What it does
The server holds a database of ready-made 1C code templates (over 2200 from the community) and gives an agent six MCP tools: hybrid semantic and full-text search, a paginated list, fetching a template by id, adding, updating and deleting. Data is stored in SQLite as the source of truth and indexed in ChromaDB for vector search; embeddings can come from an OpenAI-compatible API (LM Studio, Ollama) or be computed locally with a SentenceTransformer model, with automatic fallback if the API is unavailable. A full web interface lets you edit templates with BSL highlighting via Monaco Editor. A separate, sizable companion reference for the 1C:ZUP 3.1 API lives in docs/rules and can be attached to an agent as extra context.
Who it is for. For 1C developers who want an agent to reuse vetted code templates and find them by meaning rather than by exact name.
Good fit when
- You need a large ready BSL template database for common tasks, not an empty collection
- You want to search for a template by the meaning of a Russian-language question
- You work with 1C:ZUP and want to attach an agent the companion HR API reference
Not a fit when
- You do not want to run a Docker stack with ChromaDB and an embedding model
- Exact name search is enough without semantics
- You need to edit a template directly in the database without a later export to seed_templates.jsonl: skipping that step loses edits on rebuild
Example request
Find the average earnings calculation template for ZUP and show the code with its descriptionLimitations
There is no license file in the repository. Edits made through the web interface live only in the Docker volume runtime database and are lost on docker compose up --build or volume removal unless exported to seed_templates.jsonl beforehand. The RoSBERTa model for better Russian-language quality needs a GPU or 5-10 minutes of CPU indexing. The GPU profile requires the nvidia-container-toolkit to be installed.
How to disable. Stop the container with docker compose down and remove the server from your MCP client config.
MCP
- Transport
- http
- Authentication
- not required
| Environment variables | |
|---|---|
| EMBEDDING_PROVIDER | Embedding provider: auto, local or openai. |
| OPENAI_API_BASE | URL of the OpenAI-compatible embeddings API, for example a local LM Studio instance. |
| OPENAI_API_KEY secret | Key for the OpenAI-compatible API, not needed for local LM Studio by default. |
Security check
- Only works with the local template database, does not connect to a live 1C base
- The MCP endpoint and web interface are open with no authorization on the chosen port
README in short
The README describes installing from a ready Docker image or building from source with cpu and gpu profiles, a table of six MCP tools, web interface routes, a full list of config variables, an architecture diagram (FastAPI, SQLite, ChromaDB, embeddings), three embedding provider modes, the process of adding a template via JSONL or the web UI with a data-loss warning, and a separate section on the 1C:ZUP companion reference with setup instructions for Claude Code and Cursor.
FAQ
Is an external embedding service required?
No, by default the auto mode tries an OpenAI-compatible API and falls back to a local model if it is unavailable.
How do you avoid losing a template added via the web UI?
Export the runtime database to seed_templates.jsonl with the export_to_jsonl.py script before rebuilding the image or removing the volume.
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