AnyPDF - Your PDF Converter

Operation - Compress an uploaded PDF (by storage key) and return a downloadable link. Prerequisites - On ChatGPT with native attachments, call upload-document-native first to obtain a storage key. - On other hosts (e.g. Claude MCP), use the upload-document widget to obtain a storage key. - If a storage key already exists from a prior step, pass it directly. Inputs - fileKey: storage key of the source PDF. - level: compression intensity. Validation - fileKey must belong to the authenticated user namespace (uploads/<userId>/... or outputs/<userId>/...). Output - New PDF key + download URL + size. Use the returned key for subsequent operations. Response guidance - If the original size is known in context, mention compression ratio/percent. Do not invent sizes. Example - fileKey="uploads/abc/my.pdf", level="recommended"

Operation - Convert an uploaded document (by storage key) to another format and return a downloadable link. Prerequisites - On ChatGPT with native attachments, call upload-document-native first to obtain a storage key. - On other hosts (e.g. Claude MCP), use the upload-document widget to obtain a storage key. - If a storage key already exists from a prior step, pass it directly. Supported conversions - Images: JPG/JPEG/PNG -> PDF - Microsoft Office: DOCX/DOC, PPTX/PPT, XLSX/XLS -> PDF - OpenDocument: ODT, ODP, ODS -> PDF - Text: TXT, RTF -> PDF - PDF -> JPG (converts all pages up to 100; single-page returns JPG, multi-page returns ZIP of images) Inputs - fileKey: storage key of the source file. - conversion: source/target format pair. Validation - fileKey must belong to the authenticated user namespace (uploads/<userId>/... or outputs/<userId>/...). Output - New file key + download URL + size. Use the returned key for subsequent operations. - For multi-page PDF to JPG: returns a ZIP file containing all page images.

Operation - Extract (keep) specific pages from an uploaded PDF and return a new PDF. Prerequisites - On ChatGPT with native attachments, call upload-document-native first to obtain a storage key. - On other hosts (e.g. Claude MCP), use the upload-document widget to obtain a storage key. - If a storage key already exists from a prior step, pass it directly. Inputs - inputKey: storage key of the source PDF. - pageRanges: comma-separated page numbers/ranges (1-based). Validation - inputKey must belong to the authenticated user namespace (uploads/<userId>/... or outputs/<userId>/...). Output - New PDF key + download URL + size containing only requested pages.

Operation - Merge multiple uploaded PDFs (by storage keys) into a single PDF. Prerequisites - On ChatGPT with native attachments, call upload-document-native with all files in a single call (up to 10) to obtain storage keys. - On other hosts (e.g. Claude MCP), use the upload-document widget to obtain storage keys. - If storage keys already exist from prior steps, pass them directly. Inputs - inputKeys: ordered list of source PDF keys. Validation - each key must belong to the authenticated user namespace (uploads/<userId>/... or outputs/<userId>/...). Output - New merged PDF key + download URL + size.

Operation - Rotate every page of an uploaded PDF and return a new PDF. Prerequisites - On ChatGPT with native attachments, call upload-document-native first to obtain a storage key. - On other hosts (e.g. Claude MCP), use the upload-document widget to obtain a storage key. - If a storage key already exists from a prior step, pass it directly. Inputs - inputKey: storage key of the source PDF. - angle: clockwise rotation angle (90/180/270). Validation - inputKey must belong to the authenticated user namespace (uploads/<userId>/... or outputs/<userId>/...). Output - New rotated PDF key + download URL + size.

Interactive document upload widget for selecting files from the user's device. When to call - The user wants to perform a PDF/document operation but has NO file attached in the composer. - On hosts without native file attachments (e.g. Claude Desktop, other MCP clients), ALWAYS use this widget to obtain files. - As a fallback when upload-document-native is unavailable or fails. When NOT to call - A usable storage key (uploads/... or outputs/...) already exists from a previous step. - On ChatGPT Apps SDK, if the user attached files in the composer, prefer upload-document-native first. Workflow 1. User requests an operation (e.g. "Compress this PDF"). 2. No storage key is available in context. 3. Invoke this tool — the upload interface is displayed. 4. User selects and uploads file(s). 5. Storage keys are returned via widget state or follow-up message. 6. Use the returned keys with the appropriate operation tool. Supported file types - PDF, JPG/JPEG, PNG - Microsoft Office: DOC, DOCX, PPT, PPTX, XLS, XLSX - OpenDocument: ODT, ODS, ODP - Text: TXT, RTF

Upload one or more files that the user attached in the chat composer into app storage. When to call - The user attached files in the ChatGPT composer and asks for any PDF/document operation. - ALWAYS call this tool FIRST, before any operation tool (compress-pdf, convert-document, merge-pdfs, extract-pages, rotate-pdf). - Pass all attached files in the files array. When NOT to call - Storage keys (uploads/... or outputs/...) are already available from a previous step — pass them directly to the operation tool. - The host does not support native file attachments (e.g. Claude Desktop, other MCP clients). Use the upload-document widget instead. - NEVER fabricate or guess file arguments. Only use file objects provided by the host. - Do NOT pass local file paths such as /mnt/data/... as file arguments. Output - Returns one storage key per file (uploads/<userId>/...) to use as input in downstream operation tools. Supported file types - PDF (.pdf) - Images: JPG, JPEG, PNG - Microsoft Office: DOC, DOCX, PPT, PPTX, XLS, XLSX - OpenDocument: ODT, ODS, ODP - Text: TXT, RTF

Critical - TOOL - NEVER INVOKE THIS DIRECTLY. Internal-only backend workflow. - GPT/LLM must NEVER call this tool in user workflows. Use upload-document for user uploads. Operation - Generate presigned Cloudflare R2 upload URLs for a batch of files. Inputs - files: list of file descriptors (id, filename, contentType, contentLength). Output - Presigned upload URLs + storage keys for each file.