Proposal
Add a new crate to expose bux as an MCP (Model Context Protocol) server, enabling AI assistants (e.g., Claude) to execute code in isolated micro-VMs.
Motivation
AI assistants need safe, isolated environments for code execution. bux provides strong isolation via micro-VMs, making it ideal for this use case. MCP integration would allow seamless integration with AI tools without requiring users to manage VM lifecycle manually.
Proposed Tools
The MCP server should expose the following tools:
- create_vm: Create a VM from an OCI image
- exec: Execute commands in a VM
- write_file: Write files to a VM
- read_file: Read files from a VM
- stop_vm: Stop a VM
- list_vms: List all VMs
Implementation
Create with:
- MCP server implementation (using official MCP SDK)
- Tool handlers wrapping bux Runtime API
- Configuration for data directory and payload paths
- Optional: resource limits and timeout controls
Alternatives
- Use bux-serve HTTP API with custom MCP bridge (more complex)
- Direct CLI invocation via MCP (less efficient, no state management)
Additional Context
MCP specification: https://modelcontextprotocol.io
This would enable use cases like:
- AI code execution in sandboxed environments
- Safe testing of untrusted code
- Isolated dependency management for AI workflows
Proposal
Add a new crate to expose bux as an MCP (Model Context Protocol) server, enabling AI assistants (e.g., Claude) to execute code in isolated micro-VMs.
Motivation
AI assistants need safe, isolated environments for code execution. bux provides strong isolation via micro-VMs, making it ideal for this use case. MCP integration would allow seamless integration with AI tools without requiring users to manage VM lifecycle manually.
Proposed Tools
The MCP server should expose the following tools:
Implementation
Create with:
Alternatives
Additional Context
MCP specification: https://modelcontextprotocol.io
This would enable use cases like: