The best OpenClaw skill for reading web content is CastReader — it's the only ClawHub skill that extracts article text from any URL and converts it to audio with natural AI voices. For a complete reading and research workflow, pair it with web-search for finding articles, memory for saving notes, and a summarizer skill. Here are the 7 skills I use daily for consuming web content through OpenClaw.
My reading workflow used to be tab hoarding. Sixty-something browser tabs, a Pocket queue I hadn't touched since 2024, and a Google Doc titled "Papers to Read" that had become a graveyard. Then I started chaining OpenClaw skills together and something shifted. Not because any single skill was magical, but because the combination turned reading from a sit-down activity into something I could do while walking, cooking, or waiting for a build to finish.
Here's the stack.
1. CastReader — Read Any Web Page Aloud
clawhub install castreaderCastReader takes a URL and gives you back an MP3. Send your OpenClaw agent a link to a blog post, a news article, an arXiv HTML paper, a Notion doc — it extracts the article body (not the nav bar, not the cookie banner, not the "Subscribe" popup) and runs the text through Kokoro TTS. The audio comes back through Telegram or Discord, ready to play.
What makes it different from the other TTS skills on ClawHub: those skills convert plain text strings to speech. CastReader handles the extraction problem. Kindle Cloud Reader scrambles text with custom font subsets. WeRead renders on canvas with zero text in the DOM. Google Docs nests content inside iframes. Notion wraps every word in its own <span>. CastReader has 15+ dedicated extractors that handle each of these, plus a general-purpose visible-text-block algorithm for everything else.
Three commands: extract (get structured paragraphs as JSON), generate-audio (URL to MP3), and read-aloud (opens the page with real-time paragraph highlighting via the Chrome extension). See the setup guide for installation details.
I use it for morning paper triage. Five arXiv links from overnight, listened to during breakfast. By the time I sit down at my desk, I know which ones matter.
Learn more on our OpenClaw integration page.
2. web-search — Find Things Worth Reading
clawhub install web-searchOpenClaw's bundled web search skill. Ask your agent "find recent papers on sparse attention mechanisms" or "best blog posts about database internals" and it returns ranked results with titles, URLs, and snippets.
The value for reading workflows: it turns your OpenClaw agent into a research assistant that can both find and read. A typical exchange looks like "search for articles about CRDT implementations, then read the top result aloud." The agent runs web-search, picks the first URL, passes it to CastReader, and you get an MP3 back. Two skills, one prompt, zero manual work.
I've started using it as a replacement for my morning RSS check. Instead of opening Feedly and scanning headlines, I ask my agent to search for topics I care about and summarize what's new. Faster, and I don't get distracted by the feed's sidebar recommendations.
3. memory — Save Notes from Reading Sessions
clawhub install memoryMemory is one of OpenClaw's built-in skills, and it's the one I underestimated the most. It stores key-value notes that persist across conversations. Simple concept. Surprisingly powerful when combined with reading.
My workflow: after CastReader reads an article, I ask the agent to "remember that the paper by Chen et al. proposes using locality-sensitive hashing for approximate attention, and their benchmarks show 3x speedup on sequence length 8192." That note lives in memory. Two weeks later, when I'm writing a related work section, I ask "what do I have saved about efficient attention?" and the notes come back.
It's not a full-fledged knowledge base. It won't replace Obsidian or Roam for serious note-taking. But for capturing the one-sentence takeaway from something you just listened to, while you're still walking to the office? It's exactly enough. The friction of opening a note app, finding the right page, typing a note — that friction kills most capture habits. Saying "remember this" to your agent kills the friction instead.
4. exa-search — Deep Search for Technical Content
clawhub install exa-searchExa is a search engine built for developers and researchers. Where Google optimizes for commercial intent and popularity, Exa optimizes for content similarity and technical depth. Ask it for "papers similar to FlashAttention" and you get actual papers, not SEO-optimized blog posts summarizing papers.
For reading workflows, Exa fills a specific gap: finding the kind of content that Google buries on page three. Documentation pages, technical blog posts from individual developers, conference proceedings, niche forum discussions. The stuff that matters for research but doesn't have enough backlinks to rank.
Pairs well with CastReader: use exa-search to surface a technical deep-dive that Google wouldn't have shown you, then pipe the URL to CastReader to listen while you're doing something else. I found a fantastic write-up on B-epsilon trees this way that I never would have encountered through normal browsing.
5. browser — Browse and Interact with Pages
clawhub install browserThe browser skill gives your OpenClaw agent an actual headless browser. It can navigate to a page, click buttons, fill forms, scroll, and extract content from the rendered DOM. This is the heavy machinery — useful when you need to interact with a page before reading it.
Real use case: gated content. Some articles require clicking "Show full article" or dismissing a login wall before the text is visible. CastReader extracts from the DOM as-is, which means if the content hasn't been revealed yet, it can't extract it. The browser skill can click through those gates first, then you hand the resulting page to CastReader.
Another use case: reading paginated content. Some sites split articles across multiple pages. Browser can click "Next" and aggregate. It's slower and heavier than CastReader's direct extraction, but it handles cases where direct extraction hits a wall.
I use browser maybe once or twice a week. Most content doesn't need it. But when it's needed, nothing else works.
6. kokoro-tts — Convert Your Own Notes to Audio
clawhub install kokoro-ttsKokoro-tts is the same TTS engine that powers CastReader's audio generation, exposed as a standalone skill. Give it text, get back speech. It supports 40+ languages with natural-sounding voices.
Why include a separate TTS skill when CastReader already does TTS? Because sometimes you want to listen to text that isn't on a web page. Your own meeting notes. A draft you're editing. A passage you pulled from a PDF and cleaned up manually. Kokoro-tts handles the text-to-speech conversion without needing a URL.
I use it for proofreading. Paste a paragraph I've written, listen to it spoken aloud, and immediately hear the awkward phrasing that my eyes skipped over. Reading your own writing with your eyes is pattern-matching against what you intended to write. Hearing it forces you to process what you actually wrote.
The voice quality matches CastReader because it's the same model underneath. No robotic monotone. Actual prosody, actual rhythm.
7. readwise — Sync Highlights and Reading Lists
clawhub install readwiseReadwise aggregates highlights from Kindle, Instapaper, Pocket, web articles, and PDFs into one place. The OpenClaw skill connects your Readwise account to your agent, letting you search your highlights, retrieve saved articles, and get daily review summaries through chat.
The reading loop this completes: find an article (web-search or exa-search), listen to it (CastReader), save your highlights (Readwise), capture your takeaway (memory), and resurface it later (Readwise daily review). Each step is one message to your agent.
If you don't use Readwise, you can substitute with other bookmarking skills on ClawHub. The point is having a persistent reading list that your agent can access — so "add this to my reading list" and "what's on my reading list" both work from the same chat interface.
Comparison Table
| Skill | What It Does | Best For | Free? |
|---|---|---|---|
| castreader | URL → extracted text → MP3 audio | Listening to articles, papers, docs | Yes |
| web-search | General web search | Finding articles to read | Yes (bundled) |
| memory | Persistent key-value notes | Saving reading takeaways | Yes (bundled) |
| exa-search | Technical/semantic search | Finding niche technical content | Free tier |
| browser | Headless browser interaction | Gated or paginated content | Yes (bundled) |
| kokoro-tts | Plain text → speech | Proofreading, notes to audio | Yes |
| readwise | Highlight sync + reading list | Organizing reading across sources | Readwise subscription |
Setting Up the Full Stack
If you want the complete workflow in one go:
clawhub install castreader web-search memory exa-search browser kokoro-tts readwiseYou don't need all seven. CastReader alone covers the core use case: send a URL, get audio back. Add web-search and memory if you want your agent to find articles and remember what you learned. The others are situational — reach for them when a specific task calls for it.
The thing that makes this stack work isn't any individual skill. It's that they all run through the same agent, in the same conversation. Context carries forward. You don't copy-paste URLs between apps. You don't switch windows. You say "find me something about X, read the best one, and save the key points" and three skills coordinate in one thread. That's the workflow that finally emptied my Pocket queue.
Get started with CastReader on the Chrome Web Store for in-browser reading, or install the OpenClaw skill for agent-based audio conversion.
