Agents are good at deciding what to do and bad at physically doing it. Rebind closes that gap: a script can run a loopback HTTP server, and local software can press keys and move the mouse. WebSocket listeners are network-visible and require authentication and firewall restrictions. With Rebind Link attached, that input arrives as standard USB HID from real hardware.
Use case
Your agent gets hands.
Let a model or a program drive a real keyboard and mouse over HTTP or WebSocket. Rebind is the hardware hand for software agents.
use cases / ai agents
How it fits together
- A Rebind script exposes an endpoint on your machine.
- Your agent, in any language, sends it commands: press this, type that, move here.
- The engine emits the input. The computer sees a keyboard and a mouse.
Typed clients for TypeScript, Python, and Rust speak the protocol directly, so an agent loop is a few lines of ordinary code. In hardware mode, Rebind Link emits standard USB HID. Current hardware specifications live on the Rebind Link product page.
What people build
A CI job that drives a GUI-only installer through its screens.
An LLM that operates legacy desktop software no API ever existed for.
A Python process that plays back recorded input choreography for demos.
A local workflow trigger: a request arrives, a macro runs at the desk.
RebindGPT itself, generating the scripts the agent will run.
The full protocol, client libraries, and examples live on the agentic control page and in the documentation. Or start by generating a script in RebindGPT.