1C MCP Suite: server bundle for a 1C configuration
1C MCP Suite
Five independent Docker-based MCP servers: a Neo4j metadata and call graph, static BSL analysis, platform syntax help, a query builder and 1C coding standards
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Indexers read the whole 1C configuration code and metadata into their Neo4j and Qdrant stores
- The hidden metadata_reload and metadata_cypher tools run arbitrary graph operations if explicitly enabled by a flag
Install
Manual install
docker compose build && docker compose up -dBefore starting, fill .env with MCP_SHARED_SECRET and NEO4J_PASSWORD, otherwise the servers exit with code 78.
This is third-party code. Review the repository files before installing.
What it does
The suite brings up five MCP servers over Streamable HTTP, each with its own storage. Metadata-graph builds a Neo4j graph of the configuration structure and procedure call graph from an XML export and BSL code, answering questions like who calls this function or which objects belong to no subsystem. Bsl-checker wraps the BSL Language Server and checks code or a directory against configured rules. Platform-help does hybrid search (dense vectors plus BM25) over the official platform help shipped as .hbk files, via Qdrant. Query-builder reads the same metadata graph and helps build, validate and parse 1C query language statements. The fifth server, v8std-mcp, is not written by the project's authors; it is pulled in from the third-party zeegin/v8std repository with 1C coding standards. Every server the project owns returns answerable and degraded fields in its responses so an agent does not mistake an empty result for a broken tool.
Who it is for. For 1C developers and teams who need an agent to understand a specific configuration: its objects, relations, code and standards compliance.
Good fit when
- You need the agent to have a call and metadata graph for a specific configuration, not just general reference
- You want to run static analysis on 1C code straight from an agent conversation
- You need local offline search over the platform syntax helper and coding standards
Not a fit when
- You do not have Docker Desktop with at least 8 GB of memory allowed
- You are not ready for the first indexing run: help indexing takes an hour to ninety minutes on CPU
- You need one simple MCP server, not infrastructure with five containers plus Neo4j and Qdrant
Example request
Find every procedure that works with the ЗаказПокупателя document and show whether any of them are dead code with no incoming callsLimitations
The first image build takes 10-15 minutes, and the first help index build takes an hour to ninety minutes on CPU. Servers refuse to start without MCP_SHARED_SECRET and NEO4J_PASSWORD set. The only transport is Streamable HTTP, no SSE. The public variant of v8std-mcp sends code snippets to an external ai.v8std.ru address, which is why a local run with its own standards corpus is the default. You need an XML configuration export and .hbk help files; without them there is little point deploying this. Windows delivers filesystem events through the bind mount with a delay, so automatic graph updates are off by default.
How to disable. Stop the suite with docker compose down, and add the -v flag to also remove the indexed data.
MCP
- Transport
- http
- Authentication
- API key
| Environment variables | |
|---|---|
| MCP_SHARED_SECRET required, secret | Shared secret for bearer auth on the project's own servers; without it they refuse to start |
| NEO4J_PASSWORD required, secret | Neo4j database password, only applied on the volume's first initialization |
| WATCHER_ENABLED | Enables automatic graph updates when an export file is saved |
Security check
- Indexers read the whole 1C configuration code and metadata into their Neo4j and Qdrant stores
- The hidden metadata_reload and metadata_cypher tools run arbitrary graph operations if explicitly enabled by a flag
README in short
The README, over 800 lines, describes the five-server lineup with ports and tool counts, Docker memory requirements, step-by-step secret setup and environment checks, how each server and its tools work, the difference between full and partial reindexing, everyday docker compose commands, and an extensive troubleshooting section with concrete symptoms like exit code 78 or a 503 caused by a Windows system proxy.
FAQ
Why a separate server just for static analysis?
Bsl-checker is the only one with no storage or state: it wraps the BSL Language Server and always answers with a fresh check, independent of whether the metadata graph is ready.
Can I use the public v8std instead of the local one?
Yes, with one line in the client config pointing at https://ai.v8std.ru/mcp, but then code snippets are sent to a third-party server, which is not suitable for every project.
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