CLI

Roadmap

Planned OmniScout features — design stubs for upcoming releases.

This page describes features not yet implemented. See the commands reference for what ships today.

Multi-provider search

Today search uses DuckDuckGo (--source ddg) plus optional local indexes.

Planned:

omniscout search "query" --provider brave
omniscout search "query" --provider auto   # first provider with a configured API key

Candidate providers: Brave, Exa, Tavily, SearxNG.

Smart extraction

Today omniscout extract @eN resolves a ref to the active page URL and fetches readable content.

Planned natural-language targets:

omniscout extract "pricing table"
omniscout extract "main article"
omniscout extract "documentation examples"

Resolution would combine snapshot ref scoring with DOM heuristics (tables, article, code blocks).

Page summaries

omniscout summarize              # current page
omniscout summarize --extract  # last extracted content
omniscout summarize --search     # latest search results

Would reuse the existing sumy-based summarization stack from research.

MCP server (deferred)

No MCP server ships today. Agents use the CLI (--json) and the local daemon at 127.0.0.1:7720.

A future MCP surface would expose tools such as search, open, snapshot, extract, click, and fill over stdio for Claude Desktop, Cursor, and other MCP hosts.

Edit this pageorReport an issue

Architecture

OmniScout's daemon, backends, snapshot model, and storage layout.

Troubleshooting

Common OmniScout failure modes and how to fix them — especially when an AI agent is driving.