# How Many Files Can You Upload to Claude? Chat vs Projects Limits If you are asking how many files can I upload to Claude, standard chat accepts up to 20 files per message at up to 500MB each, while Claude Projects imposes no fixed file-count cap for documents up to 30MB each. Instead, total project knowledge is bounded by Claude's 200,000-token context window. Here is how chat attachments and persistent project files differ, what happens when you exceed capacity, and how teams manage larger document collections. Source: https://fast.io/resources/how-many-files-can-i-upload-to-claude/ Author: [Tom Langridge](https://fast.io/authors/tom-langridge/) Last reviewed: 2026-10-27 ## How Many Files Can You Upload to Claude in Standard Chat? According to [Anthropic's official support documentation](https://support.claude.com/en/articles/8241126-upload-files-to-claude) verified in September 2026, standard Claude chat accepts up to 20 files per message at up to 500MB each, while Claude Projects allows unlimited files up to 30MB each within the context window. When you attach files directly to a message in Claude, the interface allows up to `20` separate documents or images in a single prompt submission. This 20-file threshold applies uniformly across the web interface, official desktop application, and mobile apps. A widespread point of confusion stems from conflicting online search results. Many guides cite a 5-file cap or a 50-file limit. Standard Claude chat accepts up to `20` files per prompt, alongside separate boundaries for project-level storage. Guides asserting a 5-file limit do not reflect current limits. | Claude Environment | Maximum File Count | Per-File Size Limit | Processing Constraints | Verified Date | |---|---|---|---|---| | Standard Chat (Free & Paid) | Up to 20 files per message | 500MB per file | Subject to model context window (200,000 tokens) | September 2026 | | Claude Projects (Pro & Team) | Unlimited files (no fixed cap) | 30MB per file | Bounded by project context ceiling (200,000 tokens) | September 2026 | | Claude Desktop App | Up to 20 files per message | 500MB per file | Subject to local memory and model context window | September 2026 | | Claude Files API (Developers) | Unlimited files per organization | 500MB per file | 1 TB total organization storage limit | September 2026 | | PDF Documents (Chat & Projects) | Counts toward file cap | 500MB chat / 30MB project | 100 pages for visual analysis; 1,000 pages text-only | September 2026 | | Image Uploads (JPEG, PNG, GIF, WebP) | Counts toward chat file cap (20 files) | 500MB per file | Up to 8,000 x 8,000 pixels; downscaled if larger | September 2026 | Claude accepts standard documents including PDF (.pdf), Microsoft Word (.docx), CSV (.csv), plain text (.txt), HTML (.html), OpenDocument Text (.odt), Rich Text Format (.rtf), and EPUB (.epub). Microsoft Excel spreadsheets (.xlsx) are supported once code execution and file creation are enabled in account settings. Supported image formats include JPEG (.jpg, .jpeg), PNG (.png), GIF (.gif), and WebP (.webp). Audio, video, and executable packages cannot be uploaded directly to chat. ### File Size and Dimension Thresholds for Chat Attachments While Claude permits individual files up to `500MB` in standard chat conversations, this storage ceiling is designed for data-heavy assets like high-resolution graphics and scanned PDF documents. Claude cannot process `500MB` of uncompressed text in a single prompt. Because text is converted into tokens during processing, a dense text file exceeding a few megabytes will immediately exhaust Claude's context window. Image uploads carry dimension boundaries alongside file weight limits. Images can measure up to `8000x8000` pixels. When an uploaded image exceeds these dimensions, Claude downscales the image to fit operational resolution thresholds while preserving original aspect ratios. For optimal optical character recognition, images should be at least `1000x1000` pixels. PDF processing follows a strict three-tier architecture based on page counts: 1. Documents containing `100` pages or fewer receive full multimodal processing, analyzing embedded diagrams, flowcharts, styling, and structural layouts alongside text. 2. Documents spanning `101` to `1000` pages are processed through plain-text extraction only, disregarding visual graphics. 3. Documents exceeding `1000` pages are rejected immediately by the upload parser, returning an explicit file error. To analyze documents exceeding `1000` pages, you must divide the file into smaller volumes before submitting it to the chat interface. ## Does Claude Projects Have a File Count Limit? Claude Projects has no fixed file count cap. You can upload dozens or hundreds of files into a project knowledge base without encountering a hard count rejection. Anthropic's official support documentation confirms that file count in Projects is unlimited, provided that every individual file remains under `30MB` and cumulative extracted text fits within Claude's 200,000-token context window. The genuine constraint governing Claude Projects is token capacity rather than file quantity. When you upload reference files to a project, those documents form background knowledge for every conversation inside that workspace. Claude injects this extracted text directly into the prompt context. Understanding token arithmetic explains why project capacity varies across document types: - In English text, one token corresponds to roughly 4 characters or 0.75 words. - A concise 1-page memo of 500 words consumes roughly 650 to 700 tokens. - A dense 50-page technical whitepaper with tables can consume 35,000 to 45,000 tokens. - A collection of multiple comprehensive PDF manuals can total hundreds of thousands of tokens, exceeding Claude's active context capacity entirely. Consequently, you could upload numerous compact code files or markdown notes that occupy only a fraction of available context. Conversely, uploading four heavy financial audits could saturate the project knowledge meter immediately. Anthropic provides a graphical capacity meter inside the project interface displaying the proportion of context consumed by your reference files. ### Chat Attachments Compared to Persistent Project Knowledge Choosing between chat attachments and project knowledge depends on how your team reuses documentation across sessions: - **Persistence Across Sessions**: Files attached to a chat message remain bound strictly to that discussion thread. Starting a new conversation requires re-uploading files. Files uploaded to a Project's Files tab persist across all conversations opened within that project. - **Per-File Size Limits**: Standard chat allows individual files up to `500MB`, whereas Claude Projects enforces a ceiling of `30MB` per file. An `80MB` handbook uploads directly to a chat message but fails with a size error in a Project. - **Multimodal Interpretation**: Chat attachments support multimodal evaluation for PDFs of `100` pages or fewer and standalone images. In Claude Projects, files are processed through text extraction only, with the exception of multimodal PDFs. Non-PDF files containing embedded diagrams have their visual components omitted. - **Context Allocation per Interaction**: When you attach files to a chat message, they occupy context only in that specific discussion. In a Project, every uploaded document is loaded into the model's context for every prompt. If project files consume 140,000 tokens, every message starts with only 60,000 tokens of remaining headspace for ongoing discussion history and generated responses. Teams using Claude Pro or Claude Team plans must balance project breadth against response depth. Overloading a project with peripheral documentation restricts the working memory available for complex reasoning. ## What Happens When You Exceed Claude Upload and Context Limits? When you exceed Claude's upload boundaries, the system reacts with specific error states depending on whether you breached a file count cap, a file size threshold, a page limit, or total token capacity: - **Exceeding the 20-File Chat Cap**: Selecting more than `20` files in your operating system's file browser and dragging them into the chat prompt causes the web interface to reject the batch. The attachment tray either disables the surplus files or presents an error notification indicating that message attachments cannot exceed 20 files. You must deselect extra files or submit them across consecutive messages. - **Exceeding the Per-File Size Ceiling**: Dropping a file larger than `500MB` into a chat prompt or a file larger than `30MB` into a Project triggers an immediate banner stating that the uploaded file is too large. Transfer halts before server-side parsing begins. - **Exceeding the 1,000-Page PDF Boundary**: If a PDF document contains more than `1000` pages, Claude rejects the upload outright with an upload failure notification, even when the file size on disk is small. - **Context Window Saturation**: When total document text approaches the 200,000-token threshold, Claude displays an alert stating that the conversation is too long. In standard chat, the interface prompts you to start a new conversation. In Claude Projects, reaching full capacity on the project knowledge meter blocks the creation of new chats until you remove existing files. Beyond explicit error dialogues, operating near maximum token capacity introduces qualitative performance issues. Large language models experience retrieval degradation when searching for specific details within massive context payloads, a phenomenon known in machine learning research as the needle-in-a-haystack problem. If a project contains heavy volumes of technical documentation, Claude may overlook specific clauses buried in the middle of long files. Because Claude reserves tokens for reasoning and output generation, a saturated context window restricts the length and depth of the model's answers. Technical file formatting issues can also mimic capacity failures: - **Password Protection**: Encrypted PDF and Office files produce an extraction error. Decrypt files before uploading. - **Scanned Text Without an OCR Layer**: If a PDF consists of raster image scans without an embedded text layer, Claude's text extraction pipeline may extract minimal content, resulting in empty answers. - **Spreadsheet Configuration**: Attempting to upload .xlsx files while code execution and file creation remain disabled in account preferences causes file rejection. - **Damaged Metadata**: Corrupted file headers or invalid character encodings in plain-text and CSV files can halt parsing during ingestion. ### Troubleshooting Steps for Failed File Ingestion When an upload fails unexpectedly, follow this diagnostic sequence: 1. **Verify File Weight and Page Counts**: Check file properties on your local filesystem. Confirm that chat attachments do not exceed `500MB` and project files do not exceed `30MB`. For PDF documents, open file properties in your PDF reader and confirm the total page count is `1000` or fewer. 2. **Inspect Attachment Batch Counts**: Confirm that your current prompt contains no more than `20` attached items. If you have already attached 18 files and attempt to drag 4 more, the surplus will be refused. 3. **Audit Token Density**: If a text file fails to upload, check whether it contains repetitive logs or dense tabular data. High-density text files can overwhelm token limits despite small disk footprints. Split the file into focused chapters or extract only pertinent sections. 4. **Export Clean File Versions**: For spreadsheets and complex Word documents, re-export the file as clean CSV or Markdown. Stripping away legacy XML structures, macros, and embedded media resolves parsing conflicts. ## How to Manage Large Document Collections in Claude Workflows Engineering teams, research groups, and operational departments frequently work with documentation libraries that dwarf Claude's native upload allowances. When your knowledge base comprises hundreds of architectural specs, thousands of customer contracts, or gigabytes of product manuals, relying on `20`-file message batches or `30MB` project uploads becomes unworkable. Teams employ three primary strategies to handle large document collections. The first strategy is manual pre-processing and aggregation. Practitioners use command-line scripts or document utilities to split long manuals into modular chapters, remove non-essential columns from bulky CSV exports, and strip decorative formatting from corporate documentation. While this approach keeps individual files under `30MB` and `1000` pages, it imposes ongoing maintenance overhead. Every time an internal policy or software specification updates, someone must manually re-slice the source document and re-upload it to Claude. The second strategy involves rotating project files dynamically. A team working on legal discovery or medical research might maintain a primary document archive on local disk, uploading only the few files relevant to the active task into a Claude Project. Once that research phase concludes, the user deletes those files from the project and uploads the next batch. While this preserves context window headroom, it prevents Claude from synthesizing insights across different project phases and destroys institutional memory across team members. The third and most sustainable approach decouples document storage from Claude's context window entirely using external workspaces connected through the Model Context Protocol (MCP). Instead of forcing dozens of large files through Claude's web upload dialogue, organizations store their full document corpus in an intelligent cloud workspace such as [Fast.io workspaces](/product/workspaces/). In this architecture, documents live in shared, organization-owned workspaces. The files are indexed on arrival, making their contents searchable by meaning and metadata. When an operator asks Claude a question, the assistant does not read the entire document collection at once. Instead, Claude queries the workspace over an MCP connection, retrieves only the specific passages relevant to the query, and loads those excerpts into the active prompt. This retrieval-augmented architecture offers several operational benefits: - **Zero Context Waste**: Your document archive can span many gigabytes across thousands of files, yet Claude consumes only a few thousand tokens to ingest the precise paragraphs needed to answer a prompt. The remainder of the 200,000-token context window remains available for reasoning, multi-turn dialogue, and code generation. - **Elimination of File Size Limits**: Files in external workspaces are not bound by Claude's `30MB` project ceiling or 20-file message cap. Single files can measure in gigabytes without triggering ingestion errors. - **Workspace Knowledge Synchronization**: Documents stay synchronized on a schedule or on demand with primary repositories rather than growing stale inside an isolated chat project. Teams can explore this pattern at [Fast.io storage for agents](/storage-for-agents/). ### Connecting Claude to External Workspaces via MCP Connecting Claude to an external workspace using the Model Context Protocol requires no complex local software compilation. Fast.io provides a remote MCP server hosted over Streamable HTTP at `https://mcp.fast.io/mcp/tools`. This endpoint connects Claude directly to shared workspaces where documents are indexed for retrieval. Complete setup documentation is maintained at [Fast.io MCP docs](https://mcp.fast.io/docs). A Fast.io workspace serves as a shared, organization-owned environment where files are indexed upon upload. Cloud Sync connects Google Drive, Dropbox, Box and OneDrive folders to a workspace, one-way or two-way, on a schedule or on demand. Teams can also use URL Import to pull documents directly from cloud storage without routing files through local disk input and output. Once files arrive in a workspace, Intelligence Mode indexes document text for hybrid search, combining full-text keyword indexing, semantic search, and metadata value filtering. For structured extraction, [Metadata Views](/product/document-data-extraction/) turn collections of PDFs, scanned forms, and spreadsheets into queryable tabular databases with typed fields, allowing Claude to inspect extracted columns without reprocessing raw files. To connect Claude to your Fast.io workspace, follow these steps in the Claude interface: 1. In Claude on the web, the desktop app, or mobile, open **Customize**, then select **Connectors**. 2. Select the option to add a custom connector and enter `https://mcp.fast.io/mcp/tools` as the connector URL. 3. Select **Connect**. A browser window will open displaying the Fast.io sign-in and authorization screen. 4. Review permissions. Choose between Read Only or Read & Write access, and select which organizations and workspaces the connector can reach. 5. In any chat conversation, turn Fast.io on from the **+** menu under Connectors. Readers never need to edit local configuration files like `claude_desktop_config.json` to connect Fast.io. The connector operates over Streamable HTTP with OAuth authentication. Once enabled, Claude can search workspace documentation, inspect per-file version history, and review collaborative notes directly from the chat prompt. Workspaces also maintain a detailed activity log of changes, recording uploads, edits, moves, and deletions made by both team members and connected AI assistants. Fast.io monthly plans start with a 14-day free trial, which requires a credit card. Paid subscription plans are Starter at `$9.99/mo`, Business at `$49.99/mo`, and Enterprise at `$199.99/mo`. Details are available on the [Fast.io pricing page](/pricing/). Maximum single-file upload sizes are `25 GB` on Starter, `50 GB` on Business, and `100 GB` on Enterprise, far exceeding standard AI chat upload boundaries. ## How Do Claude File Limits Compare Across Plans and Interfaces? Claude's file upload rules interact differently with the account tiers and client interfaces Anthropic provides. Understanding these plan nuances ensures that you choose the right environment for your file analysis tasks. ### Free Claude Accounts Free accounts on the Claude web interface allow up to `20` files per message at up to `500MB` per file. Free accounts do not include access to Claude Projects. All file uploads are strictly conversational and ephemeral. Free accounts operate under dynamic rolling rate limits that reset every few hours. Attaching multiple large files consumes substantial compute during document parsing, which can cause free users to reach their message allowance after only a few interactions. ### Claude Pro Claude Pro is Anthropic's individual subscription tier. Pro users retain the standard 20-file and `500MB` per-message limits in chat, while gaining full access to Claude Projects. Pro subscribers can create multiple distinct projects, uploading an unlimited number of files up to `30MB` each into project knowledge bases within the 200,000-token context ceiling. Pro accounts also provide higher message volume allowances compared to unpaid tiers, enabling sustained analytical work across multi-file document sets. ### Claude Team and Enterprise Claude Team and Enterprise plans are designed for organizational collaboration. These plans provide all the capabilities of Claude Pro, augmented with collaborative project sharing. Team members can contribute to shared project knowledge repositories, ensuring that everyone references identical documentation. Administrative consoles allow workspace owners to manage member access and review billing centrally. Enterprise tiers also offer higher context usage limits and expanded operational capacity. ### Claude Desktop Application The official Claude Desktop application for macOS and Windows mirrors the web interface's 20-file and `500MB` upload limits. While the desktop client offers a convenient native interface, it processes files through the same cloud-based ingestion pipeline. Uploading a document in the desktop app transmits the file to Anthropic's cloud infrastructure for parsing and tokenization; it does not process documents locally on your machine. ### Claude Code Claude Code is Anthropic's terminal-based agent tool designed for software engineering workflows. Rather than attaching files through a graphical file picker, developers point Claude Code at local repository directories. Claude Code inspects source trees, reads files on demand, and executes terminal commands directly. While it bypasses the 20-file chat attachment dialogue, it remains bounded by model context windows. Claude Code manages this boundary using context compaction, subagent delegation, and targeted file inspection rather than loading entire repositories into active memory. ### Anthropic Files API For software developers building custom applications, Anthropic provides the Files API in beta. The Files API allows developers to upload documents up to `500MB` each and store up to `1 TB` of total file volume per organization. Files uploaded via the API receive unique identifiers that can be referenced across multiple API calls, eliminating the need to re-transmit large file payloads with every prompt request. This server-side persistence provides programmatic developers with capabilities analogous to Claude Projects, but managed entirely through software endpoints. ### Selection Checklist for Feeding Documents to Claude When deciding how to structure your document workflows in Claude, use this decision framework: - **Use Standard Chat Attachments** when analyzing a one-off document, comparing a batch of distinct receipts or invoices, or conducting an exploratory analysis where files will not be referenced again in future sessions. - **Use Claude Projects** when working on an ongoing initiative that requires persistent access to a curated library of reference manuals, style guides, or project briefs totaling well within the context window ceiling. - **Use the Files API** when developing proprietary software applications that require server-side file persistence and automated prompt construction across repeated customer requests. - **Use External Workspace Storage with MCP** when your documentation library exceeds `30MB` per file, spans hundreds of individual documents, or needs to stay synchronized on a schedule or on demand with Google Drive, Dropbox, Box, or OneDrive without exhausting Claude's context window. Discover dedicated options at [Fast.io AI storage](/product/ai-storage/). ## Frequently asked questions ### How many files can I upload to Claude at one time? Standard Claude chat allows uploading up to 20 files per message across both free and paid accounts. Each file can weigh up to 500MB, provided the extracted text fits within Claude's 200,000-token context window. ### Does Claude Projects have a file count limit? Claude Projects has no fixed file count limit. You can upload an unlimited number of files, provided each individual file does not exceed 30MB and the cumulative volume of all uploaded files fits within the project context ceiling. ### What is the maximum file size you can upload to Claude? In standard chat conversations, the maximum file size is 500MB per file. In Claude Projects, the maximum file size is 30MB per file. In addition, PDF documents are limited to a maximum of 1,000 pages. ### What happens when you exceed Claude's upload capacity? If you attempt to attach more than `20` files in a chat prompt, the file picker will block the extra files. If an individual file exceeds 500MB in chat or 30MB in Projects, Claude displays an 'Uploaded file is too large' error. If total document text exceeds the 200,000-token context window, Claude displays a 'Conversation is too long' notice. ### Can you upload folders directly to Claude? The Claude web interface does not support uploading uncompressed folders directly. You must select individual files or multi-select up to `20` files at once. To work with structured folder hierarchies, connect an external workspace via Model Context Protocol. ### Do PDF page counts affect Claude upload limits? Yes. PDFs of 100 pages or fewer receive full visual and textual analysis. PDFs between 101 and 1,000 pages are processed for plain text only. PDFs exceeding 1,000 pages are rejected outright by the system. ### How can you give Claude access to hundreds of documents? Because Claude Projects caps files at 30MB and is bounded by a 200,000-token context window, large archives cannot be uploaded directly. Instead, store documents in an external cloud workspace, enable automatic indexing, and connect Claude via MCP using [Fast.io storage for agents](/storage-for-agents/) so Claude can query indexed files dynamically. ## Sources - [Claude Help Center: Upload files to Claude](https://support.claude.com/en/articles/8241126-upload-files-to-claude): Standard Claude chat accepts up to 20 files per message at up to 500MB each, while Claude Projects allows unlimited files up to 30MB each within the context window. ## About Fast.io Fast.io provides shared workspaces where people and AI agents work on the same files, with built-in semantic search and citation-backed chat over what they hold. Agents reach it through a remote MCP server, a REST API at https://api.fast.io/current/, and a command line client published on npm as @vividengine/fastio-cli. MCP setup is at https://mcp.fast.io/docs: Claude and most MCP clients connect to https://mcp.fast.io/mcp/tools, ChatGPT to https://mcp.fast.io/mcp/operations, and coding agents to https://mcp.fast.io/mcp/code.