Eversale CLI

Your AI Employee - Desktop agent for research, sales, admin, and automation.

Quick Install

curl -fsSL https://eversale.io/install.sh | bash

Or via npm:

npm install -g eversale-cli

Getting Started

1. Run the CLI - It will prompt for login:

   eversale

2. Enter your email - A magic link will be sent

3. Enter token from email - Activates your license

4. Start automating!

   eversale "Research Stripe competitors"

Requirements

• Node.js 18+ - For the CLI wrapper
• Python 3.10+ - For the AI engine
• License - Purchase at https://eversale.io/desktop

Install Python (if needed)

macOS:

brew install [email protected]

Ubuntu/Debian:

sudo apt install python3.12 python3.12-venv

Usage

Interactive Mode

eversale

One-Shot Mode

eversale "Research Stripe competitors"
eversale "Find 10 marketing agencies in Miami"
eversale "Scrape product prices from amazon.com"

Commands

eversale                    # Interactive mode (prompts for login if needed)
eversale "task"             # One-shot task execution
eversale login <email>      # Send login email
eversale activate <token>   # Activate with token from email
eversale logout             # Remove license
eversale --setup            # Re-run setup
eversale --help             # Show help
eversale examples           # Show example prompts

Speed Mode (High-Speed for Internal Tools)

For processing high-volume tasks on internal/safe websites, use natural language to activate speed mode for 40-150x performance:

# Just say it naturally!
eversale "fast track process 1000 items on the dashboard"
eversale "quickly fill out all the forms"
eversale "turbo mode - export all leads to CSV"
eversale "rush through the admin panel updates"

# All these phrases work:
# fast track, quickly, turbo, speed mode, instant, rapid, asap, rush, go fast

Performance: | Operation | Normal | FAST_TRACK | Speedup | |-----------|--------|------------|---------| | Mouse move | 150-400ms | <10ms | 40x | | Type 50 chars | 8-12s | <100ms | 100x | | 1000 forms | 3-4 hours | 2-3 min | 100x |

Safety: FAST_TRACK automatically rejects public sites (Amazon, LinkedIn, etc.) to prevent bot detection. Only use on:

• ✅ localhost, private IPs (192.168.x.x)
• ✅ Internal domains (*.company.local)
• ✅ Admin panels behind auth
• ❌ Never on public e-commerce/social media

How It Works

1. CLI validates your license against eversale.io
2. AI requests are routed through eversale.io to our GPU servers (RunPod)
3. Browser automation runs locally on your machine
4. Results are saved to ~/.eversale/outputs/

Request Flow

Your Machine                     Cloud
    ↓                              ↓
eversale CLI → eversale.io → GPU Server (RunPod)
    ↓                              ↓
Playwright ←────────────────── LLM Response
    ↓
Browser actions run locally

Configuration

Configuration is stored at ~/.eversale/engine/config/config.yaml

Key settings:

• llm.mode - "remote" (default, uses eversale.io) or "local" (requires Ollama)
• browser.headless_default - Run browser hidden (default: false)

Using Your Own GPU Endpoint (Pod / Serverless)

The engine supports alternative backends via either:

• EVERSALE_LLM_URL env var (always overrides), or

• llm.allow_custom_endpoint: true + llm.base_url: ... in ~/.eversale/engine/config/config.yaml.

• OpenAI-compatible (vLLM/TGI/etc): set llm.base_url to the server root (no /v1 needed). Example: http://<host>:8000

• Ollama-compatible: http://localhost:11434

• RunPod Serverless: set llm.base_url to https://api.runpod.ai/v2/<endpoint_id> and set RUNPOD_API_KEY in your environment

Notes on speed:

• Serverless often has cold start latency; to match pod-like speeds, configure your endpoint with a warm worker / minimum workers (provider setting).

Troubleshooting

Having issues? Just tell the agent what's wrong:

eversale "check my setup"           # Runs health checks
eversale setup                      # Re-runs installation

Common Issues

| Problem | Solution | |---------|----------| | Can't login | Check your email for the login link, or run eversale again | | Browser not working | Run eversale setup to reinstall | | Something broken | Run eversale setup to fix it |

Need help? Email [email protected] or visit https://eversale.io/desktop

Related Repositories

| Repo | Path | Description | |------|------|-------------| | ev29 | /mnt/c/ev29/ | Monorepo (web app, CLI, agent backend) | | cli | /mnt/c/ev29/cli/ | This CLI package (published to npm) | | agent-backend | /mnt/c/ev29/agent-backend/ | Python servers/workflow runtime; the CLI engine is the single brain (agent-backend shims into cli/engine/agent) |

Infrastructure

| Server | Purpose | |--------|---------| | eversale.io | Web app, license validation, LLM proxy | | RunPod GPU | LLM inference (RTX A5000) | | PostgreSQL | Database |

Support

• Website: https://eversale.io/desktop
• Dashboard: https://eversale.io/desktop/success?login=1
• Email: [email protected]