1C AI sandbox client-server
1c-ai-sandbox-client-server
A devcontainer plus a Hyper-V VM with a 1C server that isolate an AI agent from the host machine so a wrong command cannot touch real files or drives
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
- Deploys a virtual machine and Docker containers with full administrator rights on the host
- Scripts handle real 1C platform license data if DEV_LOGIN and DEV_PASSWORD are set
Install
Manual install
pwsh -NoProfile -ExecutionPolicy Bypass -File .\scripts\tests\SmokeTest-OnecInfra.ps1Deploys the server infrastructure on Hyper-V from an administrator PowerShell.
This is third-party code. Review the repository files before installing.
What it does
The project sets up two isolated parts for safely letting an AI agent work with 1C. The client part is a Linux dev container with 1C thick and thin clients, including the Configurator, a headless Xvfb display and wrapping for the Claude, Codex, Gemini and DeepSeek Harness CLI agents with ready bootstrap scripts. The server part is a separate Ubuntu VM in Hyper-V with a Docker Compose stack of a 1C server, PostgreSQL and Apache for web publications, deployed by a PowerShell script that downloads and patches the Ubuntu ISO automatically. The idea is that the agent works only inside the container and the VM and physically cannot delete or change anything on the host.
Who it is for. For 1C developers who want to give an AI agent real access to the 1C platform but worry a wrong command could touch a working drive or a production infobase.
Good fit when
- You want to give an agent a real 1C client and server, but isolated from the host machine
- You have a Windows host with Hyper-V and Docker Desktop to deploy the server VM
- You work in Cursor or VS Code with Dev Containers
Not a fit when
- You do not have a Windows host with Hyper-V: the server scenario is only verified there, Linux hosts are untested
- You need a lightweight setup without a virtual machine and Docker
Example request
Deploy the sandbox, bring up the demo test infobase and publish it through the web clientLimitations
The server scenario is only verified on a Windows host with Hyper-V and PowerShell run as administrator, Linux hosts with KVM, QEMU or VirtualBox are untested. Per the author, community license auto-activation is currently broken on the platform side, so the DEV_LOGIN and DEV_PASSWORD variables should be left empty. Deployment requires Docker Desktop, Hyper-V and downloading the official Ubuntu ISO.
How to disable. Remove the dev container from VS Code or Cursor and delete the virtual machine in Hyper-V Manager if it is no longer needed.
Security check
- Deploys a virtual machine and Docker containers with full administrator rights on the host
- Scripts handle real 1C platform license data if DEV_LOGIN and DEV_PASSWORD are set
README in short
The README explains the project's motivation with a story of losing personal files to an agent's mistake, and describes the architecture of a client dev container and a server Hyper-V VM with a connection diagram. It gives step by step instructions for preparing .env files, running the server-part deployment script and publishing infobases through Apache, plus explicit limitation warnings: tested only on a Windows host.
FAQ
Why build a sandbox like this at all?
The author lost 10 years of family photos to an AI agent's mistake and built an isolated environment so the agent physically cannot delete anything outside the container and VM.
Which CLI agents are already set up?
.devcontainer/cli-agents has ready bootstrap scripts for Claude, Codex, Gemini and DeepSeek.
Related
An MCP server built into the Netdata agent: metrics, logs, alerts and live process, service and container data for an AI assistant
GitHub's official MCP server: code, issues, pull requests, Actions and security alerts straight from the agent
Agent Skills for Google products
Agent Skills for Google products and technologies
Official Google skill collection for working with Google Cloud, BigQuery, GKE, ads and analytics from an agent
AWS's official MCP server suite: docs, IaC, containers, serverless, databases, cost and monitoring