Arrakis MCP Server
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., "@Arrakis MCP Server Create a sandbox and run 'whoami'"
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.
Arrakis MCP Server
A Model Context Protocol (MCP) server that exposes Arrakis VM sandbox functionality to Large Language Models.
Setup
Setup
Clone this repository:
git clone https://github.com/abshkbh/arrakis-mcp-server.git cd arrakis-mcp-serverConfigure your LLM tool to use the Arrakis MCP server by adding the following to your MCP configuration:
"mcpServers": { "arrakis": { "command": "<path-to-uv>", "args": [ "--directory", "<path-to-repo>", "run", "arrakis_mcp_server.py" ] } }
Related MCP server: agent-sandbox-mcp
API
The Arrakis MCP Server exposes the following MCP resources and tools:
Resources
arrakis://vms- List all available VMsarrakis://vm/{vm_name}- Get information about a specific VM
Tools
start_sandbox- Start a new VM sandboxrestore_snapshot- Restore a VM from a snapshotsnapshot- Create a snapshot of a VMrun_command- Run a command in a VMupload_file- Upload a file to a VMdownload_file- Download a file from a VMdestroy_vm- Destroy a specific VMdestroy_all_vms- Destroy all VMsupdate_vm_state- Update the state of a VM (pause/stop)
License
This project is licensed under the GNU Affero General Public License v3.0 - see the LICENSE file for details.
Available Tools
9 toolsdestroy_all_vmsC
Destroy all VMs.
Returns: Success/failure message.
Inputs: no named parameters.
destroy_vmB
Destroy a VM.
Args: vm_name: Name of the VM to destroy.
Returns: Success/failure message.
Inputs: vm_name (required).
download_fileB
Download a file from a VM.
Args: vm_name: Name of the VM to download from. path: Path of the file to download.
Returns: Content of the file or error message.
Inputs: vm_name (required), path (required).
restore_snapshotC
Restore a VM from a snapshot.
Args: vm_name: Name to give to the restored VM. snapshot_id: ID of the snapshot to restore from.
Returns: Information about the restored VM.
Inputs: vm_name (required), snapshot_id (required).
run_commandB
Run a command in a VM.
Args: vm_name: Name of the VM to run the command in. cmd: Command to execute.
Returns: Command output and/or error.
Inputs: vm_name (required), cmd (required).
snapshotA
Create a snapshot of a VM.
Args: vm_name: Name of the VM to snapshot. snapshot_id: Optional ID for the snapshot. If not provided, one will be generated.
Returns: ID of the created snapshot.
Inputs: vm_name (required), snapshot_id.
start_sandboxB
Start a new sandbox VM.
Args: name: Name to give to the new VM.
Returns: Information about the created VM.
Inputs: name (required).
update_vm_stateA
Update the state of a VM.
Args: vm_name: Name of the VM to update. status: New state for the VM. Must be either 'stopped' or 'paused'.
Returns: Success/failure message.
Inputs: vm_name (required), status (required).
upload_fileB
Upload a file to a VM.
Args: vm_name: Name of the VM to upload to. path: Destination path in the VM. content: Content of the file.
Returns: Success/failure message.
Inputs: vm_name (required), path (required), content (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 9 tools
Each tool has a clearly distinct purpose: starting, destroying, snapshotting, restoring, file operations, command execution, and state updates. No two tools overlap in functionality.
Most tools follow a consistent verb_noun pattern (e.g., destroy_vm, download_file, restore_snapshot). The tool 'snapshot' is a slight outlier as it uses a noun as a verb, but it is still clear.
With 9 tools covering creation, destruction, state management, snapshots, file transfer, and command execution, the count is well-scoped for a VM management server.
There are significant gaps: no tool to list VMs or snapshots, and no way to get VM status. These missing query operations will cause agent failures when trying to discover existing resources.
Maintenance
Related MCP Connectors
Provides capabilities that let LLM agents perform a range of infrastructure management tasks.
Secure Linux sandboxes for AI agents, as MCP tools.
Cloud Linux VMs for your agent: run code, Docker, files and HTTPS previews in a Sandboxer microVM.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
Related MCP Servers
- AlicenseAqualityCmaintenanceConnects AI agents to microsandbox for creating lightweight sandboxes, executing code, managing files, and monitoring resources.19695 npm20Apache 2.0
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to create, manage, and execute code in isolated Firecracker microVM sandboxes via the MCP protocol, with support for sandbox lifecycle and file operations.4Apache 2.0
- AlicenseNot gradedqualityCmaintenanceProvides a restricted Docker-based sandbox for LLM agents, enabling file operations, command execution, and local Git within an isolated runtime.19 npmMIT
- AlicenseAqualityBmaintenanceProvides agents with real Linux VMs to run commands, manage files, expose ports, and track costs, all through an ephemeral sandbox service.11MIT