Working with Files, PDFs, and Long Documents

What You Can Upload to Claude

Claude.ai accepts file uploads directly in the conversation. You can upload:

• Documents — PDFs, Word (.docx), and plain text files
• Spreadsheets — Excel (.xlsx) and CSV files
• Images — JPEG, PNG, and other common image formats
• Code files — Any programming language in plain text format

Upload a file by clicking the paperclip icon in the message input or dragging and dropping it into the chat. Once uploaded, the file is part of the conversation context — Claude can read it and reference it in every response in that conversation.

Analyzing Documents

The most common document task is analysis: extracting information, summarizing, finding patterns, or answering specific questions about the content.

Effective document analysis prompts:

I've uploaded a contract. Identify any clauses that might be unusual or unfavorable to the party signing it, and explain each one in plain language.

This is a 40-page research report. Give me a 300-word executive summary focused on the key findings and recommendations.

Analyze this spreadsheet of sales data. What are the top three trends you notice, and are there any anomalies I should investigate?

Specificity matters more with documents than with any other input. "Summarize this" produces a general summary. "Summarize the key findings in the methodology section and identify any limitations the authors acknowledge" produces a targeted, useful analysis.

Working with Long Documents

Claude can handle long documents — books, lengthy reports, legal agreements — but the context window has limits. For most users, the practical limit is around 150,000–200,000 tokens (roughly 100,000–150,000 words), depending on your plan.

For documents within the limit, upload the whole thing and ask your questions. For very long documents:

• Break the document into logical sections and upload one section at a time
• Ask Claude to summarize each section, then synthesize the summaries
• Use the Project knowledge base for documents you reference across multiple conversations (documents in the knowledge base are cached and do not consume your per-message context budget)

Images: What Claude Can See

Claude can read and analyze images as well as documents. Useful applications:

• Screenshot analysis — Paste a screenshot of an error message, a UI design, a chart, or a table, and ask Claude to interpret it
• Document photos — Take a photo of a handwritten note or a physical document and have Claude transcribe or analyze it
• Chart and graph reading — Upload a chart image and ask Claude to describe the trend, identify outliers, or compare values
• Design feedback — Share a mockup or screenshot and ask for specific feedback on layout, typography, or clarity

The Project Knowledge Base for Persistent Documents

If you frequently reference the same documents across multiple conversations — a style guide, a product spec, your company's standard templates — the best solution is to upload them to a Project's knowledge base rather than re-uploading them in every conversation.

Documents in the knowledge base are available in every conversation in that Project. Claude reads and indexes them, then references them when relevant. They are cached separately from the conversation context, so they do not reduce the space available for your actual messages.

To add a document to the knowledge base: open the Project settings, navigate to the Knowledge Base section, and upload your files there. Changes take effect immediately in new conversations.

Try This Now

Find a PDF you have been meaning to read — a report, a long article, an industry document — and upload it to Claude. Ask Claude to give you the three most important ideas in it and one question the document raises that it does not fully answer. Then ask a specific follow-up question about one of those ideas. Notice how Claude references the document directly rather than relying on general knowledge.