Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Connector MCP start a background task to run the tests"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
High-performance Model Context Protocol (MCP) bridge that connects AI coding agents to remote cloud VPS environments, isolated container execution sandboxes, background task multiplexing, and persistent state across multi-tasking sessions.
Connector is designed for modern AI developer agents (Claude, Gemini, Cline, Roo, Cursor, etc.) that need access to high-performance remote computing environments without overloading or cluttering local developer machines.
By connecting to a Connector MCP Server on your VPS, your AI agent can:
Execute Long-Running Tasks: Compile large codebases, run test suites, and execute background tasks that persist even if your agent disconnects.
Multi-Task Seamlessly: Open multiple workspace tabs, switch between interactive shells and local session disks, and monitor jobs simultaneously.
Isolated Sandboxing: Each project runs inside an isolated container sandbox with zero risk to the host machine.
State & Memory Continuity: Project memory graphs and ledger records ensure seamless continuity across agent sessions and generations.
Related MCP server: sidekick
Standard JSON-RPC 2.0 protocol over HTTP / Streamable SSE.
Direct execution inside containerized project sandboxes (/workspace and /work).
Full support for interactive shell commands, background daemons, and output streaming.
Run background tasks (task run <cmd>) that keep executing in the background.
Attach, inspect, and kill background tasks on demand.
Switch between multiple concurrent workspaces without losing execution context.
Dedicated local session disk per project (~/.connector-cli/sessions/<project>).
Fast offline memory cache and Knowledge Graph exploration (graph, search, node).
Cursor-style fast AST symbol and code indexing (code, symbol, outline).
Install the Connector CLI on the agent's machine:
npm install -g connector-mcp-cliSet your VPS connection credentials (provided once by your administrator):
export CONNECTOR_URL="http://your-vps-ip:3210"
export CONNECTOR_API_KEY="your-secret-api-key"Launch the interactive workspace or connect directly to a project:
# Launch interactive project picker & multitasking shell
connector-cli
# Connect directly to a specific project
connector-cli connect <project-name>Inside the interactive project workspace:
Tab VPS: Interactive remote shell inside the VPS container.
Tab Disk Session: Local project session disk, Knowledge Graph memory, and code navigation.
Background Tasks: Spawn, manage, and monitor long-running background tasks.
Tool Name | Description |
| List all available project workspaces on the VPS |
| Open a project container and set default workspace |
| Execute a shell command inside the project container |
| Spawn a background task that survives disconnects |
| Inspect output and status of a running task |
| List all running and completed background tasks |
| Terminate a running background task |
| Read a file from the container workspace |
| Write a file to the container workspace |
| List files and directories in the workspace |
| Retrieve the project's permanent Knowledge Graph |
| Search entities and relations in the Knowledge Graph |
Rootless Container Isolation: All code execution occurs within dedicated Podman container sandboxes.
Scoped Mounts: Each container only has access to its assigned project directory.
Network Boundaries: Container network isolation ensures safe execution of untrusted third-party code.
MIT License. Designed for agentic workflows and automated developer pair-programming.
This server cannot be deployed
Zero-Ops deploy of a private AI workspace to your own VPS — from your AI chat. Free and open-source.
Shared control plane for AI coding agents — tasks, memory, decisions, file locks. 12 tools.
Deploy and manage your apps, databases, storage, and scheduled jobs from your AI agent
Build Apps and run code in 30 languages — sandboxed, with persistent sessions for agent loops.
