What is OmniScout?

OmniScout — local-first browser control, semantic search, and research for AI agents.

OmniScout is a self-contained, terminal-native interface for the web: a local browser brain for AI agents. It searches the web, extracts structured data, controls a real browser, automates desktop apps, and remembers everything — without leaving your laptop.

How it fits together

CLI — the omniscout command you (or your agent) type. Every command speaks JSON.
Browser — stays ready so actions run fast.
Search & answers — web search, page extraction, research, and knowledge graphs.

Install

pip install omniscout
omniscout install --skill

Full setup — browsers, daemon, models — is covered in Installation. The 5-minute tour is Quickstart.

What you can do

Area	Commands	Guide
Q&A with citations	`answer`	Search & answer
Ranked web hits	`search`, `map`	Search & answer
Page → Markdown/JSON	`extract`	Extraction
Multi-source research	`research`, `graph`	Research & graphs
Browser control	`browser`, `open`, `snapshot`	Browser automation
Desktop automation (macOS)	`computer`	Computer
Background service	`daemon`, `session`, `profile`	Daemon & sessions
Personal web memory	`remember`, `memory`, `index`	Memory
Reusable flows	`record`, `macro`, `replay`, `workflow`, `monitor`	Workflows
Setup & tuning	`install`, `setup`, `settings`, `warmup`, `benchmark`	Configuration

Design principles

Local-first — no cloud account or API key. Everything stays on your machine; the only outbound traffic is searches you run and URLs you fetch.
Agent-ready — --json everywhere, stable @eN element refs, and skill files for Claude Code, Cursor, Codex, and Gemini CLI.
Real browser — your Chromium, your cookies, your extensions. Log in once, reuse forever.

Use

# Start the daemon (once)
omniscout daemon start

# Search the web
omniscout answer "what is the capital of france"

# Extract content from a URL
omniscout extract https://example.com

# Control a browser
omniscout browser navigate https://example.com
omniscout browser snapshot --refs-only
omniscout browser click '@e3'
omniscout browser screenshot --out /tmp/page.png

# Research a topic
omniscout research "AI agents in 2026"

# Build a knowledge graph
omniscout graph "Cursor"

Next steps

Installation — requirements, browsers, daemon, models
Quickstart — first run in five minutes
Using with AI agents — drop-in prompts for Claude Code, Cursor, Codex
Commands reference — every verb, every flag
Examples & recipes — common workflows