Get past the Claude Projects knowledge limit.
A Claude Project fills up by size, not by file count, so a few large documents can max out its knowledge. Drop the whole folder here instead. Everything, even the PDFs, is read in your browser and packed into one file, with the token count shown so you know it fits.
- See the token count before you add it
- PDFs and Office docs read in-browser
- No sign-up, nothing uploaded
Drag your folder here
Everything for your Project, read in a second.
Claude counts size, not files.
Project knowledge shares Claude's context window, so a project is full when the total size runs out, not when you reach some number of files. That makes the real question how many tokens your documents come to. FileConcat answers it before you add anything.
Counted locally, with the same kind of tokenizer the models use
Where Claude stops you.
The caps that matter are about size and project count, not a number of files. One combined file makes the size easy to see and keeps the whole set in a single place.
| Where you add files | What it caps | Limit today |
|---|---|---|
| Project knowledge | Total size, shared with the context window | Around 200,000 tokens, not a file count |
| Projects | How many projects you can create | 5 on the free plan, more on paid |
| Chat attachments | Files on a single message | A few at a time, each size-limited |
Figures as of July 2026, and Anthropic changes them often. Check the current numbers in the Anthropic help center. On ChatGPT, Gemini, or NotebookLM instead? See how to share all your files with any AI.
From a full folder to one project file.
- 1
Drop the whole folder
Drag in every document you want the project to know about. Subfolders come along, and you can add a repo or a link too.
- 2
It reads and counts
PDFs, Word, Excel, and notes become text in this tab, and the token count is shown as it goes. Nothing is uploaded.
- 3
Add the one file to your project
Upload the single file to the project once you have seen it fits, and every chat in the project can read the whole set.
A full folder, packed and counted.
A folder that would fill the knowledge on its own goes in. One document comes out, labeled as documents, with the token count already known so you can see it lands in the window.
acme-due-diligence/
├── contracts/
│ ├── msa.pdf
│ └── sow.docx
├── financials.xlsx
├── board-notes.md
└── 27 more files<documents project="acme-due-diligence">
<summary>
This is a packed snapshot of a set
of documents, assembled by
fileconcat.com.
File count: 31.
</summary>
...One file, the whole project
A single file keeps every document together, in order, under a file tree, so the project reads it as one labeled set and every chat inside the project can draw on the whole thing.
Common questions.
- How many files can I add to a Claude Project?
- Claude does not cap a project by file count. Project knowledge shares Claude's context window, around 200,000 tokens, so what matters is the total size rather than the number of files. Combining everything into one file lets you see the token count up front and know it fits.
- Why does Claude say my project knowledge is full?
- A Claude Project fills up by size, not by file count, and it shows how much of its knowledge capacity is used. Combining your documents into one file, with boilerplate left out, keeps the whole set well under the limit.
- Does FileConcat show how much of the context window my files use?
- Yes. It counts the tokens as it combines, using the same kind of tokenizer the models use, so you see whether the whole set fits before you add it to the project.
- Can Claude read the PDFs and Word documents I combine?
- Yes. FileConcat pulls the text out of PDF, Word, Excel, and PowerPoint files in your browser and writes it into one plain-text file, which Claude reads directly without you converting anything first.
- Are my documents uploaded to a server?
- No. Every file, including the PDFs and Office documents, is read in your browser tab. Nothing is sent, nothing is stored, and there is no account to create.