Step 1
Choose the page that contains the source material you actually want Claude to reference.
Guide
Prepare Web Pages for Claude Projects
Claude Projects are most useful when the project knowledge is deliberate. A cleaner Markdown-style source note gives Claude a more readable reference while giving you a file you can inspect and revise.
Turn page dumps into source notesPreserve the URL and title for later verificationKeep long-running AI workspaces easier to trust
Core idea
The goal is source notes, not page dumps.
A project knowledge file should help Claude retrieve the right facts later. That means the source should have a clear title, useful headings, preserved links, and enough context to stand on its own. It should not begin with a menu, cookie notice, or a giant block of page chrome.
Workflow
A cleanup workflow for Claude project knowledge.
This is a simple staging process: clean the source, inspect it, add a small header, and only then move it into the workspace that will depend on it later.
Step 2
Convert the page or copied HTML into Markdown with Paepae Stack.
Step 3
Review the output and remove any remaining irrelevant sections.
Step 4
Add a short source header with URL, title, and date captured.
Step 5
Save the cleaned text as a source note.
Step 6
Add or paste that source note into your Claude project workflow.
Source note template
Give each cleaned page a small framing layer before it joins the project.
That extra note helps future-you understand why the source belongs there, not just where it came from.
Example source note
# Source Note: [Page Title]
Source URL: [url]
Captured: [date]
Prepared with: Paepae Stack HTML to Markdown for AI
## Why this source matters
[One or two sentences about how this source should be used in the project.]
## Cleaned source
[Paste cleaned Markdown here]Quality checks
Good Claude project sources are small enough to understand and strong enough to reuse.
The goal is not maximum volume. It is a set of sources whose structure and purpose are obvious when you revisit the project later.
Check 1
The title says what the source is.
Check 2
The first paragraph explains why the source belongs in the project.
Check 3
The content is organized under useful headings.
Check 4
The source URL is preserved for checking later.
Check 5
Long pages are split into topic-specific notes when needed.
Check 6
Older sources are dated so they can be refreshed later.
When to split
One long page does not have to become one long source note.
Split a source when one page covers several jobs. A long docs page might include setup, authentication, API reference, examples, and limits. If your project only needs one part, make a smaller note. Claude usually works better with deliberate context than with a giant file whose useful section is buried in the middle.
Common questions
Answer the source-note questions before you fill the project with context.
These are the questions that usually decide whether a page becomes a reliable project source note or a noisy dump that is harder to trust later.
Should I upload raw HTML to Claude Projects?
Usually no. Raw HTML can work when the markup itself matters, but most project knowledge benefits from cleaner text or Markdown-style structure.
Should every source note include the original URL?
Yes. The URL gives you a way to verify, refresh, or replace the source later.
Is this only for Claude?
No. The same workflow can help any long-running AI workspace where source quality matters. Claude Projects are simply a concrete use case.
Next steps
Use this Claude-specific branch to choose the next practical move.
From here, the best next step is to clean the source, return to the broader chatbot workflow, or compare cleaned handoff notes with raw URL-based context.
Clean the source note first
Open the tool when you want to turn a web page or copied HTML into a cleaner Markdown-style note before it enters Claude Projects.
Open HTML to Markdown for AIReturn to the broader chatbot workflow
Use the ChatGPT and Claude guide when you want the copy-paste cleanup logic without the longer-lived project-knowledge framing.
Read the chatbot guideUse the coding-workflow cousin
Move into the Cursor guide when the same source-note pattern needs to support explicit coding context instead of Claude project knowledge.
Read the Cursor guideCompare URL fetches against cleaned handoff notes
Read the URL-versus-Markdown guide when you want the deeper explanation of why cleaned payloads often travel better than raw fetches.
Read URL vs MarkdownStart from the broader cleanup category
Step back to the clean-HTML guide when you want the category-level definition of what makes source material more model-ready.
Read Clean HTML for LLMs