Core-MCP-Middleware
A JSON-RPC middleware specification between an LLM agent and a 1C HTTP service: the protocol, authentication and a fix for UUID byte order
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
- The spec describes createDocument, updateDocument, deleteDocument tools that, once implemented, would change 1C data
- There is no implementation, so the real risk depends on how a developer builds out the spec
Install
Manual install
Read architecture.md and json_rpc_spec.md from the kazancevaaroslava0-oss/core-mcp-middleware repo as reference material for the protocol and the convert_uuid_to_1c_format function; the server implementation has to be written yourself.
This is third-party code. Review the repository files before installing.
What it does
The repository is a design document, not working code: it describes the architecture of a Python and FastAPI middleware that would accept JSON-RPC 2.0 calls from an LLM agent, check JWT authentication and forward requests to a 1C HTTP service. It separately works through a specific integration problem: 1C stores UUIDs in Little-Endian byte order while Python and most LLM tooling use Big-Endian, and gives the conversion function's code. The json_rpc_spec.md file specifies three methods: Auth to get a token, GetMetadata to read an object's metadata, and ExecuteTool to call tools like createDocument, getDocument, updateDocument, deleteDocument and getCatalog, with request, response and error code examples.
Who it is for. For 1C developers designing their own JSON-RPC bridge between an LLM agent and a 1C HTTP service, who want a ready fix for the UUID byte-order problem and a protocol example.
Good fit when
- You need a JSON-RPC protocol example for agent-to-1C integration with Auth, GetMetadata and ExecuteTool methods
- You need to solve the byte-order problem when passing UUIDs between Python and 1C
- You want an architecture plan before writing your own code
Not a fit when
- You need a ready-to-run server: the repo has no implementation, only design documents
- You need the 1C HTTP service with Auth, GetMetadata, ExecuteTool methods: only the middleware side is described, the 1C part is not included
Example request
How do I correctly convert a UUID from Python to the 1C format for a JSON-RPC request?Limitations
The repo has not one line of working code: only a README, an architecture document with a single UUID-conversion function example, and a text specification of JSON-RPC methods. There is no FastAPI server, no client, no 1C HTTP service, no tests, no releases. This is a design draft, not an installable tool.
How to disable. Nothing to disable: the repo contains no working service.
MCP
- Transport
- http
- Authentication
- API key
Security check
- The spec describes createDocument, updateDocument, deleteDocument tools that, once implemented, would change 1C data
- There is no implementation, so the real risk depends on how a developer builds out the spec
README in short
The README itself is a single description line. The substance is in architecture.md and json_rpc_spec.md: the first describes the system components (user, LLM agent, Python middleware, 1C HTTP service), the data flow, security requirements (no direct SQL, JWT, HTTPS only) and the UUID conversion function; the second gives a full JSON-RPC 2.0 protocol spec with three methods and an error code table. The license is stated in the LICENSE file.
FAQ
Can I run this middleware?
No, the repo contains only design documents: a README, an architecture diagram and a JSON-RPC spec, with no working code.
What exactly is the UUID problem?
1C stores UUIDs in Little-Endian byte order while Python and most LLM tooling use Big-Endian; the document gives a function that converts the first three UUID groups.
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