summarize

About summarize

summarize is a CLI and Chrome Side Panel that turns a link, file, video, or podcast into clean text and then a sharp summary. What distinguishes it from a wrapper around a chat model is that it runs a real extraction pipeline first: Readability for articles, markitdown for files, published-transcript fetchers for media with Whisper as the fallback, and Firecrawl as a last resort when a site actively fights extraction. Only after that does it hand the cleaned content to whatever model you have wired up. Model support is deliberately broad — xAI, OpenAI, Google, Anthropic, NVIDIA, Z.AI, OpenRouter, GitHub Copilot, and Ollama for local inference, plus local CLI providers including Claude Code, Codex, Gemini, Cursor Agent, OpenClaw, OpenCode, Copilot CLI, Antigravity, and pi — so local, paid, and free models all work. Output is shaped for whatever consumes it: streamed ANSI Markdown in a terminal, plain text down a pipe, a stable JSON envelope on stdout for scripts with progress and warnings kept on stderr, and ANSI stripped under --no-color. An --extract flag skips the LLM entirely and prints just the cleaned content, which makes it useful as a plain extraction utility. For video, --slides pulls scene-change keyframes and renders them inline in Kitty or iTerm or writes them to disk. An optional daemon binds to 127.0.0.1 with a shared bearer token and pairs the CLI with the active browser tab through the Chrome extension. Defaults live in a JSON config with per-invocation flag overrides, and a local cache is on by default. MIT licensed, by Peter Steinberger.

Key Features

summarize Pros & Cons

Who Is summarize Best For?

Tags