How to Make Claude Open Links and Read the Source

Claude can fetch and read publicly accessible pages when web browsing is enabled. Here is how to ask it to open a link, confirm it read the content, handle pages it cannot reach, and save sources across sessions.

Updated 5 min read

You can ask Claude to open a link and read its content when web browsing is enabled in your app and plan. Claude can then summarize the page, extract specific information, or check a claim against it. Tools like fenbs store the recurring source links and project context that should travel with them, so important references are available in future sessions.

This guide explains when Claude can open a link, what it can do with the content, how to handle pages it cannot reach, and how to confirm it actually read what you gave it.

When Claude can open a link

Claude can open a link when web browsing is enabled and the page is publicly accessible without a login. This includes most news articles, documentation sites, blog posts, Wikipedia pages, and company websites. Paste the URL into the conversation and ask Claude to read it.

Web browsing availability depends on your plan and the app you use. In Claude.ai, look for the web search toggle in the message composer. In Claude Code and Claude Desktop, web access comes through an MCP browsing tool you configure. For a full guide to web access options, see how to give Claude internet access.

How to ask Claude to read a specific page

Paste the URL and say what you want from it. "Read this page and summarize the main argument" or "check whether this article supports the claim that X" both work well. Being specific about what you want helps Claude focus on the relevant section rather than summarizing everything.

If Claude does not automatically open the link, add a direct request: "Please fetch and read this page." Some plans use web search by default rather than URL retrieval; for a specific URL, an explicit request makes the intent clear.

Pages Claude cannot open

Claude cannot open pages that require a login, sit behind a paywall, or actively block automated access. If you paste a link and Claude says it could not access the page, one of those three is almost certainly the cause.

For pages you cannot share publicly, copy the relevant text and paste it directly into the conversation. This works for most research tasks and bypasses access restrictions without any workaround. For Reddit specifically, see why Claude cannot access Reddit.

Confirming that Claude read the page

Ask Claude to quote the passage it is drawing on. A direct quote tied to the URL you provided is evidence that it read the live page. If Claude cannot produce a quote, it may have used its training data or a cached version of the page rather than the current content.

For important claims, open the source yourself and verify. Claude can misread context, extract from an adjacent section, or apply a quote to a different point than the author intended. A quick check takes a minute and prevents an error from compounding.

Authenticated links and private resources

Some links only work when you are signed in: internal documentation, a company wiki, a shared Drive document, or a private GitHub repository. Claude cannot sign in with your credentials to access these. For private resources, copy the text and paste it, or use an MCP server configured with the right authorization.

For Google Drive links specifically, see how to give Claude access to Google Drive. Claude Code has better native support for GitHub repositories than URL fetching alone.

Saving source links for later sessions

A link Claude reads in one session is not available in the next. If a source is important to a project, save it somewhere future sessions can find it without re-fetching the page. fenbs has a Links and listings record for this: the source URL, what it showed, and any questions it raised.

Over time, a project builds up a set of trusted sources: vendor documentation, primary research, key decisions linked to their evidence. Keeping those in fenbs means any connected assistant reads the same source list at the start of a session rather than discovering them from scratch. A growing source list is also valuable when a colleague or another AI app needs to verify a claim independently. The list tells them exactly where to look without needing to search through prior chat history. Notes on what each source confirmed and what it explicitly did not address are often as valuable as the link itself, and much faster to work from than a full web page. A well-annotated source list compounds in value: each verified link makes future sessions faster to start.

How fenbs helps

Keep recurring source links with your project

Add important links to a project in fenbs alongside a note on what each source confirms. Connect Claude over MCP from Integrations and the next session reads the source list without fetching pages from scratch.

Create a free account

Free for one person · No credit card · Works with Claude, ChatGPT, Cursor and more

How to do this with fenbs

  1. Create a free account. Your personal board is free and needs no credit card.
  2. Add a project for the research or work that uses recurring source links.
  3. As you verify sources, add them to Links and listings. Note what each source confirmed and what it did not address.
  4. Connect Claude over MCP from Integrations. Ask Claude to read the project context at the start of each session.
  5. When a source changes or is superseded, update the note rather than adding a duplicate entry.

fenbs does not cache or store web pages. It keeps the source references, what they confirmed, and where the work stands so each session builds on previous findings.

Questions people ask.

Can Claude open any link I give it?

Claude can open most public pages when web browsing is enabled. It cannot open pages behind a login, a paywall, or that block automated access. Paste the content directly for those sources.

How do I know if Claude actually read the page or just guessed?

Ask it to quote the relevant passage. If it cannot produce a direct quote from the page, it may have used its training data or a cached version rather than the live content. Check important claims yourself.

Can Claude open links in a document I upload?

Not automatically. Claude can read the uploaded document and see the links, but it does not follow them on its own. You need to paste the URLs directly into the conversation and ask Claude to fetch each one.

What if the page content is behind a login?

Copy the relevant text and paste it into the conversation. Claude can work with pasted content just as well as fetched content. For private internal resources, use an MCP server that can authenticate with the service.

Can Claude save or bookmark links for me?

Claude itself does not store bookmarks between sessions. Save important sources in a record outside the chat, such as the Links and listings section of a fenbs project, so they are available in future sessions.

More in Connect your AI tools

All Connect your AI tools articles

Start with one thing.

Free for one person · No credit card · Works with Claude, ChatGPT, Cursor and more

There is nothing to set up first. Write one line and you’ve started.