1C Agent Platform: three MCP servers for 1C
1C Agent Platform
A platform of three MCP servers for reading, writing and intelligence operations on 1C:Enterprise configurations and infobases
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- mcp-write-server changes 1C configuration metadata and BSL code
- Automatic rollback covers only 6 of 25 mutating tools, worth starting with the read server
Install
Manual install
python -m mcp_read_server --transport stdioExample run of the read server over stdio; similar for the write and intelligence servers.
This is third-party code. Review the repository files before installing.
What it does
The project splits 1C work into three separate MCP servers: mcp-read-server for reads, mcp-write-server for writes, and mcp-intelligence-server for analytical operations, each with its own tool registry and console entry point. It offers stdio transport for local trusted use and a separate HTTP transport with bearer authentication via environment variables using ${ENV:NAME} substitution, rejecting a literal cleartext token at config-load time. For writes it implements a narrow rollback: 6 of 25 mutating tools are on the automatic-recovery allowlist, the rest are intentionally outside it. The README documents at length what the platform does NOT do: no RBAC, no multi-tenancy, no web UI, no universal rollback.
Who it is for. For 1C developers who need to separate read and write rights across different MCP servers and want honestly documented boundaries for automatic rollback of changes.
Good fit when
- You want to give an agent read-only access with no risk of writes
- You need to write BSL and metadata with limited automatic rollback on some operations
- You need HTTP transport with a bearer token for network access, not just local stdio
Not a fit when
- You expect a full role-based access model, in-process TLS termination or a web UI: the README explicitly says these are absent
- You need universal rollback across all mutating tools: only 6 of 25 are covered
Example request
Connect only mcp-read-server to this 1C infobase and show the sales document's metadataLimitations
The README is written in dense technical language listing development phases and tracks in detail; reading it in full is not required. The rollback whitelist covers 6 of 25 mutating tools, 24 percent of the surface, with the rest intentionally outside automatic recovery. HTTP transport is meant for a trusted network behind a reverse proxy, not the public internet directly: TLS and mTLS are not implemented inside the server.
How to disable. Stop the mcp-read-server, mcp-write-server and mcp-intelligence-server processes and remove them from your MCP client config.
MCP
- Transport
- stdio, http
- Authentication
- API key
| Environment variables | |
|---|---|
| AUTH_TOKEN_ENV secret | Name of the environment variable holding the bearer token for HTTP transport, set via --auth-token-env. |
Security check
- mcp-write-server changes 1C configuration metadata and BSL code
- Automatic rollback covers only 6 of 25 mutating tools, worth starting with the read server
README in short
The README describes 1C Agent Platform as an MCP platform for AI agents working with 1C:Enterprise configuration and infobases, split into read, write and intelligence servers. In detail, down to listing parallel development tracks, it covers stdio and HTTP transports, bearer authentication with environment variable substitution, a narrow write-rollback whitelist, and what the platform deliberately does not do: RBAC, multi-tenancy, a web UI, OS-specific packaging.
FAQ
Can I use only the read server?
Yes, the three servers are independent; mcp-read-server can be connected alone, giving the agent no write rights.
Is TLS built in?
No, HTTP transport is meant for a trusted network behind a reverse proxy; TLS and mTLS are not implemented inside the server.
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