I love local LLMs, but I hate that they are trapped in a text box. I wanted my AI to actually use my computer—open the browser, find a file, organize a folder—just like a human would.
Existing tools were mostly fragile Python wrappers or required giving the agent "god mode" (root) which terrified me.
So I built Runiq.
It’s a single static Go binary that implements the Model Context Protocol (MCP). It gives Claude/Llama access to your File System and Browser, but acts as an "OS Firewall." It intercepts every dangerous syscall (like write or delete) and forces a native popup for your approval.
Why Go?
Single Binary: No pip install dependency hell.
Concurrency: Handles the MCP stream + user input + file I/O without lag.
Safety: Stricter permission structs than Python.
It’s open source. I’d love to hear how you guys are handling agent permissions—is this overkill, or exactly what's needed?
I love local LLMs, but I hate that they are trapped in a text box. I wanted my AI to actually use my computer—open the browser, find a file, organize a folder—just like a human would.
Existing tools were mostly fragile Python wrappers or required giving the agent "god mode" (root) which terrified me.
So I built Runiq.
It’s a single static Go binary that implements the Model Context Protocol (MCP). It gives Claude/Llama access to your File System and Browser, but acts as an "OS Firewall." It intercepts every dangerous syscall (like write or delete) and forces a native popup for your approval.
Why Go?
Single Binary: No pip install dependency hell.
Concurrency: Handles the MCP stream + user input + file I/O without lag.
Safety: Stricter permission structs than Python.
It’s open source. I’d love to hear how you guys are handling agent permissions—is this overkill, or exactly what's needed?
Repo: https://github.com/qaysSE/runiq?ref=show_hn