Enro

Execution layer for robotics and AI operators on Robinhood Chain

Website · Documentation · Examples · Contributing

---

Overview

Enro is an execution layer that gives AI agents access to real machines — managed cloud, decentralized peers, GPU clusters, or your own hardware. Agents get a dedicated environment in seconds, see the screen through a vision-language operator, and interact with any application the way a person would.

No API required. No refactor required. Any screen. Any machine.

Architecture

Enro is built as three composable layers:

Layer	Crate	Description
Machines	enro-vm	Provision and manage isolated environments across cloud, peer, GPU, and local substrates
Systems	enro-core	Runtime supporting Windows, macOS, Linux, mobile emulators, and legacy stacks
Operators	enro-operator	Vision-language workers that read the screen, handle popups, and improve with use

┌──────────────────────────────────────────────────┐
│                  Enro Runtime                     │
├──────────┬──────────┬──────────┬─────────────────┤
│  Cloud   │   Peer   │   GPU    │     Local       │
│ Provider │ Provider │ Provider │    Provider     │
├──────────┴──────────┴──────────┴─────────────────┤
│              Vision-Language Operator             │
│         capture → analyze → decide → act          │
└──────────────────────────────────────────────────┘

Quick Start

# Install
cargo install enro-cli

# Start a machine
enro machine create --os windows-11 --provider cloud

# Deploy an operator
enro operator deploy --task "Extract invoice data from legacy ERP" --machine m-abc123

# Check status
enro status

Features

• Multi-substrate provisioning — Cloud, decentralized peers, GPU, or your own hardware
• Any OS — Windows 11 down to DOS, macOS, Linux, mobile emulators
• Vision-language operators — Read the screen, click, type, handle unexpected UI
• Per-agent isolation — Clean environment per task, wiped on completion
• Operator marketplace — Publish and deploy specialized operators (coming soon)
• Robinhood Chain settlement — Native metering for compute in ten-second slices

Crates

Crate	Description
enro-core	Core runtime, config, error types
enro-vm	VM provisioning and lifecycle
enro-operator	Vision and action pipeline
enro-api	HTTP API server
enro-cli	Command-line interface

Configuration

listen_addr = "0.0.0.0:8080"
data_dir = "/var/enro"
max_machines = 50

[provider]
type = "cloud"
region = "us-east-1"
api_key = "${ENRO_API_KEY}"

Contributing

See CONTRIBUTING.md for guidelines.

License

Licensed under MIT License. See SECURITY.md for vulnerability reporting.