# FileNest WorkTools > FileNest WorkTools is a browser-based productivity website for batch file renaming, image conversion and resizing, PDF/data workflows, OCR, duplicate-file checks, natural-language task routing, selected one-pass local file pipelines, and explicit agent task-completion semantics. It is not a cloud-storage or file-hosting service. ## Canonical site - English: https://filenest-pages.pages.dev/en - 中文: https://filenest-pages.pages.dev/zh - Full first-party AI context: https://filenest-pages.pages.dev/llms-full.txt - First-party problem-to-live-tool citation catalog: https://filenest-pages.pages.dev/ai-task-catalog.json (read-only; fit and not-fit boundaries; live tool and Pricing pages remain authoritative). - Tools: https://filenest-pages.pages.dev/en/tools - File workflows: https://filenest-pages.pages.dev/en/workflows - Guides: https://filenest-pages.pages.dev/en/guides - Answer hub / natural-language task router: https://filenest-pages.pages.dev/en/solutions#task-router - AI, WebMCP & browser-agent guide: https://filenest-pages.pages.dev/en/agents - Agent outcome contracts: https://filenest-pages.pages.dev/en/agent-outcomes - Natural-language intent router contract v1.0: https://filenest-pages.pages.dev/.well-known/filenest-ai-intent-router.json - AI task planner contract v1.2: https://filenest-pages.pages.dev/.well-known/filenest-ai-task-planner.json - WebMCP site-tool capability manifest v1.6: https://filenest-pages.pages.dev/.well-known/filenest-webmcp.json - Agent task contract v1.0: https://filenest-pages.pages.dev/.well-known/filenest-agent.json - Revenue-safe AI handoff contract v1.3: https://filenest-pages.pages.dev/.well-known/filenest-agent.json - Agent outcome contract v1.0: https://filenest-pages.pages.dev/.well-known/filenest-agent-outcomes.json - Workflow recipe contract v1.1: https://filenest-pages.pages.dev/.well-known/filenest-workflows.json - Benchmarks: https://filenest-pages.pages.dev/en/benchmarks ## Tools and one-pass public pipelines - [One-Pass Web Image Delivery](https://filenest-pages.pages.dev/en/tools/web-image-delivery): Add JPG, PNG or WebP once, then locally standardize format, scale, crop, quality, light adjustments and delivery filenames before one ZIP download. - [Online Batch File Renamer](https://filenest-pages.pages.dev/en/tools/batch-file-renamer): Rename files in your browser for free. Add prefixes, suffixes, text replacement, numbering and zero padding, with a before-and-after preview. - [Rename Files in Sequential Order](https://filenest-pages.pages.dev/en/tools/sequential-file-renamer): Number files in order with a custom start, increment, zero padding and number position. Preview every final filename before download. - [Reverse Number File Renamer](https://filenest-pages.pages.dev/en/tools/reverse-number-file-renamer): Number filenames in reverse for free. Customize the start, step, zero padding and file order, with a complete preview before download. - [Custom Order File Renamer](https://filenest-pages.pages.dev/en/tools/custom-order-file-renamer): Drag files into a custom order, then add prefixes, suffixes, text replacement and numbering. Mobile controls and full preview included. - [Organize Files Into ZIP Folders](https://filenest-pages.pages.dev/en/tools/batch-file-organizer): Organize mixed PDFs, images, documents and data files into ZIP folders by their original extensions. Optionally rename them, review every file, and download locally. - [Image Batch Converter](https://filenest-pages.pages.dev/en/tools/image-batch-converter): Batch convert JPG, PNG and WebP for free. Convert 5 files per 24 hours, up to 3 per batch and 25 MB each, with quota shown before processing. - [Batch Image Resize, Crop & Zoom](https://filenest-pages.pages.dev/en/tools/image-resizer): Resize images locally by percentage or exact pixels. Crop to 1:1, 4:3, 3:2, 16:9 or 9:16, adjust zoom and focus, rotate and flip. - [Batch Image Clarity & Color Enhancer](https://filenest-pages.pages.dev/en/tools/image-enhancer): Adjust sharpening, brightness, contrast and saturation in batches without uploading. Conservative defaults and honest image-quality limits included. - [Batch Image Compression & Size Optimization](https://filenest-pages.pages.dev/en/tools/compress-images-online): Batch-adjust JPG and WebP quality, optionally resize, and compare before-and-after size savings locally with ZIP download. - [Merge, Split & Create PDFs in Batches](https://filenest-pages.pages.dev/en/tools/pdf-batch-converter): Merge PDFs, split or extract pages, and turn JPG or PNG images into PDF locally. Encrypted PDFs, PDF OCR and PDF-to-Office are not supported. - [Batch CSV, TSV & JSON Conversion](https://filenest-pages.pages.dev/en/tools/csv-json-converter): Batch-convert CSV, TSV and JSON with quoted-field parsing, object-array validation, output preview and ZIP download. Data stays local. - [Text to Document Converter](https://filenest-pages.pages.dev/en/tools/text-to-document-converter): Type or paste text, choose a clean layout, built-in page background or your own image, then preview and create DOCX, PDF, TXT, Markdown or HTML locally. - [Image to Text OCR & Document Export](https://filenest-pages.pages.dev/en/tools/image-to-text-ocr): Extract text from clear JPG, PNG or WebP images. High Accuracy OCR discloses the actual route before execution: requests currently admitted to FileNest browser OCR run in a browser Worker without sending source images to external OCR, while other requests remain on the disclosed transient Azure Document Intelligence prebuilt-read route until migration completes. Local Privacy OCR remains available with no source-image upload. - [Content-Level Duplicate File Finder](https://filenest-pages.pages.dev/en/tools/duplicate-file-finder): Compare up to 500 files locally: prefilter by size, hash candidates with SHA-256, find byte-identical copies and export a CSV report. - The Web Image Delivery route is a real one-pass browser-local pipeline: the user adds JPG/PNG/WebP files once, reviews format/scale/crop/quality/light-adjustment and delivery-name settings, generates actual results, reviews bytes and warnings, then chooses whether to download one ZIP. - One-pass does not mean remote execution. The Web Image Delivery pipeline still requires browser file selection, a signed-in quota-bearing account, explicit run confirmation and a final download decision. ## File workflows - [One-pass web image delivery](https://filenest-pages.pages.dev/en/workflows#web-image-delivery): Add JPG, PNG or WebP once and finish format, scale, compression, light adjustment and delivery naming inside one local workspace. Execution mode: one_pass_local_pipeline. Current limitation: The current pipeline accepts JPG, PNG and WebP only. It is not server-side automation or a public execution API; the user still selects files, confirms settings, reviews results and decides whether to download. - [Email attachment cleanup](https://filenest-pages.pages.dev/en/workflows#email-attachment-cleanup): Reduce image attachment size first, then standardize filenames to lower attachment clutter and size-limit risk. Execution mode: guided_multi_tool. Current limitation: FileNest does not connect to your mailbox, send email automatically or read recipients. - [Scan to editable document](https://filenest-pages.pages.dev/en/workflows#scan-to-editable-document): Improve scan readability, recognize text, then export the reviewed result into a common document format. Execution mode: guided_multi_tool. Current limitation: The current OCR entry accepts JPG, PNG and WebP, not PDF. Handwriting, glare, blur and complex layouts can still produce recognition errors. - [PDF delivery pack](https://filenest-pages.pages.dev/en/workflows#pdf-delivery-pack): Clean up names and order first, then merge, split or extract the PDF pages needed for delivery. Execution mode: guided_multi_tool. Current limitation: Encrypted PDFs, PDF OCR and PDF-to-Office are not currently supported. - [Data handoff cleanup](https://filenest-pages.pages.dev/en/workflows#data-handoff-cleanup): Convert CSV, TSV or JSON into the structure the recipient needs, then standardize filenames for handoff. Execution mode: guided_multi_tool. Current limitation: FileNest does not claim support for every non-standard CSV dialect; top-level JSON must be an object or array of objects. - Machine-readable workflow contract: https://filenest-pages.pages.dev/.well-known/filenest-workflows.json. It describes current one-pass local pipelines and guided multi-tool handoffs, including privacy boundaries, limitations and review requirements. - Guided multi-tool workflows do not imply automatic cross-tool file transfer. One-pass local pipelines may combine already-supported browser operations inside one workspace, but they still do not create a public remote execution API. ## Practical guides - [How to batch rename files without uploading them](https://filenest-pages.pages.dev/en/guides/batch-rename-files-without-uploading): Learn how to batch rename files privately in your browser without uploading them. Covers prefixes, suffixes, replace, numbering, custom order, conflict checks and safe preview. - [Should batch renaming use regex, a CSV map or basic rules?](https://filenest-pages.pages.dev/en/guides/batch-rename-files-with-regex-csv-mapping): Choose regex replacement, original,new CSV filename mapping or basic batch-renaming rules with preview, conflict checks, saved templates and clear Pro boundaries. - [How to turn finished text into Word or PDF without uploading the body text](https://filenest-pages.pages.dev/en/guides/text-to-word-pdf-without-uploading): Turn typed or pasted text into DOCX, PDF, TXT, Markdown or HTML locally in your browser. Compare Word vs PDF, preview layout, use white/built-in/custom backgrounds, and understand current 24H Pass vs Pro boundaries. - [How to convert CSV, JSON and TSV locally while checking the data structure first](https://filenest-pages.pages.dev/en/guides/csv-to-json-without-uploading-data): Convert CSV→JSON and JSON→CSV/TSV locally in your browser. Understand quoted fields, object arrays, nested values, spreadsheet-formula risk, preview, batch boundaries, and when the 24H Pass or Pro fits. - [How to confirm exact duplicate files without uploading them](https://filenest-pages.pages.dev/en/guides/find-exact-duplicate-files-without-uploading): Find byte-identical duplicate files with size prefiltering and browser-local SHA-256 while distinguishing same-name, same-size, similar-looking and exact copies. - [Should you use local OCR or cloud OCR?](https://filenest-pages.pages.dev/en/guides/local-vs-cloud-ocr-privacy-accuracy): High Accuracy OCR follows the request route shown in the workspace before execution. Only requests admitted by the current release status, browser, language and device conditions use a FileNest browser Worker without sending source images to an external OCR API; other requests use the disclosed transient Azure Document Intelligence prebuilt-read cloud route. A FileNest browser request that has started does not automatically switch to external OCR after failure. This conditional routing description does not mean FileNest-owned OCR has fully replaced Azure. - [How to unmerge a PDF or split pages without Adobe](https://filenest-pages.pages.dev/en/guides/merge-split-pdf-without-uploading): Unmerge a combined PDF by selecting known page ranges. Split pages without Adobe, test a real six-page sample, and review the output. No file upload required. - [How to shrink oversized email images without uploading the originals](https://filenest-pages.pages.dev/en/guides/compress-images-for-email-without-uploading): When email photo attachments are too large, compress JPG, PNG and WebP locally in your browser. Understand dimensions, quality, format and actual byte-size tradeoffs. - [How to convert and compress images without uploading them](https://filenest-pages.pages.dev/en/guides/convert-images-without-uploading): Convert JPG, PNG and WebP, resize batches and compress images locally without uploading originals. Understand format, dimensions, quality and privacy tradeoffs. - [How to extract text from images without uploading them](https://filenest-pages.pages.dev/en/guides/ocr-images-without-uploading): Learn how to run private image-to-text OCR in your browser without uploading contracts, receipts or screenshots. Covers accuracy, review, languages and export choices. ## Question-led answer hub - [File Workflow Answer Hub](https://filenest-pages.pages.dev/en/solutions): natural-language FileNest questions mapped to a real tool, a practical guide and first-party evidence. The hub also contains a human-visible bounded task router at #task-router. - Machine-readable recommendation profile: https://filenest-pages.pages.dev/ai-recommendation-profile.json. It publishes explicit good-fit and not-fit conditions for the same supported question clusters without naming competitors or claiming an independent ranking. - Comprehensive first-party retrieval context: https://filenest-pages.pages.dev/llms-full.txt. It expands the same current registries into tools, answer-cluster fit/not-fit guidance, workflows, WebMCP boundaries, agent contracts, guide content and evidence summaries for larger-context retrieval. ## Natural-language task routing - Machine-readable bounded intent router v1.0: https://filenest-pages.pages.dev/.well-known/filenest-ai-intent-router.json. - resolve_filenest_intent accepts one short natural-language file-job description and deterministically maps it to a published FileNest task, a published workflow, or no-match. - The router does not call an external or generative model, does not store the task sentence, does not read files, does not navigate and does not prefill or process anything. Match strength is rule evidence, not a statistical confidence or accuracy percentage. - Unsupported requirements such as public remote APIs, server-side automation, cloud-drive automation, PDF-to-Office, PDF OCR, encrypted-PDF handling and destructive automation return no-match instead of being forced into a nearby FileNest tool. - open_filenest_workflow is a separate allowlisted navigation action for published workflow IDs. It does not access files, auto-transfer outputs between tools or auto-advance workflow steps. - Recommended sequence for an AI client is: resolve_filenest_intent → plan_filenest_task when a fixed Answer Hub cluster exists, or open_filenest_workflow for a workflow outcome → optional explicit Safe AI Prefill/navigation → live user review → explicit processing → explicit download. ## AI task planning - Machine-readable bounded planner: https://filenest-pages.pages.dev/.well-known/filenest-ai-task-planner.json. It maps the six current Answer Hub clusters to a real FileNest task, preferred tool, Good Fit, Not Fit, evidence, and a recommended next WebMCP action. - plan_filenest_task is read-only. It does not navigate, prefill settings, access files or execute processing. It returns no-match when the user's job cannot be accurately placed inside FileNest's fixed published fit clusters. - A matched planner result is FileNest self-description, not a cross-vendor ranking. The agent should respect not-fit conditions and use no-match instead of forcing FileNest for unsupported formats, enterprise requirements or server-side automation needs. - Planning and action are deliberately separated: plan_filenest_task can recommend open_filenest_task or an existing Safe AI Prefill tool, but a second explicit tool call is required before navigation or prefill occurs. - The csv-json-local plan can recommend prepare_data_conversion, which only prepares bounded JSON / CSV / TSV output settings. It does not infer input content or parser dialect from remote data. - The pdf-without-upload plan can recommend prepare_pdf_workflow, which only prepares current PDF merge, split/extract or images-to-PDF settings. It cannot inspect PDF contents, page counts, passwords or unsupported formats before the user selects a local file. ## WebMCP site tools - FileNest exposes browser-native WebMCP site tools on localized pages when the current browser implements the experimental document.modelContext API. Capability manifest: https://filenest-pages.pages.dev/.well-known/filenest-webmcp.json. - list_filenest_capabilities is read-only and returns a bounded FileNest task catalog, processing boundary and access requirements without reading user files. - resolve_filenest_intent is read-only deterministic natural-language routing. It calls no external model, stores no goal text, accesses no files and supports no-match. - plan_filenest_task is read-only fit planning over fixed FileNest Answer Hub clusters. It returns evidence-backed Good Fit / Not Fit guidance and supports no-match without changing page state. - open_filenest_task accepts only FileNest-published task IDs and navigates to a same-origin live page. It does not accept arbitrary URLs and does not select, read, upload, process or download files. - open_filenest_workflow accepts only FileNest-published workflow IDs and navigates to the same-origin workflow anchor. It does not access files or silently carry an output into a later tool. - prepare_batch_rename is Safe AI Prefill for the free browser-local Batch File Renamer. It validates and stores only bounded core rules such as prefix, suffix, find/replace, extension preservation and numbering in same-origin sessionStorage, then navigates to the live renamer. It deliberately excludes regex and CSV mapping from this AI-prefill contract, stores no file data, expires within one hour and never selects, reads or renames files automatically. - The Batch File Renamer visibly reports when AI-suggested rules were loaded. The user still chooses files, reviews the generated old-to-new filename list and validation warnings, and explicitly decides whether to package and download renamed copies in a ZIP. Source files are not overwritten. - prepare_data_conversion is Safe AI Prefill for the browser-local CSV / JSON converter. It validates and stores only target output format (JSON, CSV or TSV) and optional JSON pretty-print preference in same-origin sessionStorage, then navigates to the live converter. It accepts no arbitrary URL, file payload, delimiter override or parser code, stores no file data, expires within one hour and never reads, parses or converts files automatically. - The CSV / JSON converter visibly reports when AI-suggested settings were loaded. The user still selects local files, reviews the locally generated pre-conversion preview, explicitly starts conversion after quota checks, reviews generated output and decides whether to download. - prepare_pdf_workflow is Safe AI Prefill for the browser-local PDF toolkit. It validates and stores only the published merge, split/extract or images-to-PDF mode and mode-specific page range, per-page split or A4 / US Letter / match-image page-size setting in same-origin sessionStorage. It accepts no arbitrary URL, file payload or password, stores no file data, expires within one hour and never reads or processes a PDF automatically. - The PDF toolkit visibly reports when AI-suggested settings were loaded. The user still chooses local PDF or JPG/PNG files, reviews order, page counts and output plan, explicitly starts processing after quota checks and decides whether to download. Safe AI Prefill does not add encrypted-PDF, PDF OCR or PDF-to-Office support. - prepare_web_image_delivery is Safe AI Prefill for the Web Image Delivery pipeline. It validates and stores only bounded format, quality, scale, crop, light-adjustment and naming settings in same-origin sessionStorage, then navigates to the live tool. The setting record expires within one hour, stores no file data and never starts processing automatically. - The Web Image Delivery page visibly labels AI-suggested settings, lets the user change or reset them, and still requires the user to choose files, review filenames and previews, explicitly start processing, review actual outputs and decide whether to download. - get_filenest_safety_boundary is read-only and returns the user-control boundary for site tools. - WebMCP support, intent routing and Safe AI Prefill do not create a public remote MCP server or public remote file-execution API. Browsers without WebMCP continue to use the existing human web UI normally. - WebMCP site tools must not be interpreted as permission to bypass local file selection, authentication, quota, privacy disclosure, preview, processing confirmation or the final download decision. ## AI and browser-agent task handoff - [Safe agent handoff guide](https://filenest-pages.pages.dev/en/agents): current browser-task boundaries, WebMCP site-tool behavior, natural-language routing, read-only task planning, Safe AI Prefill, user-control rules, success definitions and recovery paths. - Machine-readable FileNest-owned web-task contract: https://filenest-pages.pages.dev/.well-known/filenest-agent.json. This describes evidence-reviewed public web capabilities only and is not a public remote execution API. - [Human-readable task outcome contracts](https://filenest-pages.pages.dev/en/agent-outcomes): when each published FileNest task may actually be called complete, including required context, success signals, review points, non-terminal states and recovery. - Machine-readable agent outcome contract v1.0: https://filenest-pages.pages.dev/.well-known/filenest-agent-outcomes.json. - The outcome contract is completion metadata only. It does not grant file access, remote execution, destructive permissions or permission to bypass the live page. - Routing success, opening a page, Safe AI Prefill, file selection, clicking Start, or a model/provider returning success are not sufficient by themselves to claim the user's file job is complete. - AI assistants may select the correct tool or workflow, explain prerequisites and assist with the public UI, but must not bypass browser file selection, authentication, quota checks, privacy disclosure, preview, confirmation or download decisions. - Do not treat an intent route, planner result, click, WebMCP navigation or settings/rules prefill as successful file execution. Success requires the live tool or pipeline to produce the expected result or complete the stated comparison, plus any required review and user decision described by the outcome contract. - FileNest does not authorize destructive automation. Do not delete or overwrite source files on the user's behalf. - A newly published composite pipeline may appear in the workflow contract before the separate evidence-reviewed agent-task registry is expanded. Do not invent an agent-task registration that is not present in /.well-known/filenest-agent.json. ## Revenue-safe AI commercial handoff — contract v1.3 - Machine-readable authority: https://filenest-pages.pages.dev/.well-known/filenest-agent.json. The commercialHandoffs extension is guidance for fit and navigation, not a checkout API or permission to change entitlements. - One-off or uncertain intent: tool_or_workflow → result_review → stop_without_commercial_pressure. - Recurring-work intent: repeat_work_fit → real_workflow → result_review → workload_pricing_only_after_actual_limit_or_explicit_upgrade_fit_intent. - Explicit pricing intent: live_pricing_authority → human_checkout_decision. - AI referral by itself is not commercial intent. Repeated workload, an actual Free/batch bottleneck, or an explicit pricing/upgrade question must supply the commercial reason. - A page visit, workload-fit signal, pricing view or checkout start is not revenue. Positive net revenue requires confirmed payment and positive net amount after refunds. - AI systems must not invent an acquisition source, rewrite a campaign identity, or present an observed payment association as causal attribution. - Pricing authority: https://filenest-pages.pages.dev/en/pricing. Capability authority: https://filenest-pages.pages.dev/en/workflows. Do not invent prices, discounts, trials, billing periods or capabilities from this text. - Repeated office file operations: route id=office_repeat_work; campaign-noindex handoff. Use when: The user explicitly describes recurring attachment, PDF, or data handoff work | The main pain is weekly or ongoing repetition of the same file steps rather than a one-off conversion | The user wants to verify whether current FileNest workflows can reduce repeated handoffs; do not use when: A single light file task | No recurring-work signal has been provided | Mailbox, shared-drive, or enterprise automation that FileNest does not currently provide; next decision: Verify that a current workflow completes the real job first; move to workload-pricing only when capacity or repetition becomes an actual bottleneck.. Resolve the exact URL from the machine-readable agent contract only after fit is established. - Client delivery file operations: route id=client_delivery; campaign-noindex handoff. Use when: An agency, freelancer, or small team repeatedly prepares client image, PDF, or data delivery packs | The user values pre-delivery review, source-file protection, and consistent naming | The need is the current FileNest file-handoff layer rather than a client portal or project-management platform; do not use when: A one-off personal conversion | A client portal, automatic upload, team seats, or cloud sync is required | The requested file type or operation is not supported by current FileNest workflows; next decision: Complete one safely testable real delivery batch first; evaluate Free versus Pro workload fit only after the route succeeds and repeats.. Resolve the exact URL from the machine-readable agent contract only after fit is established. - Choose Free or Pro by workload: https://filenest-pages.pages.dev/en/workload-pricing — stage=pricing_decision; use when: The user is asking whether their workload justifies upgrading to Pro rather than only asking for the current numeric price | A current FileNest route already completes the job but Free capacity, batch scale, or repetition is now a real limit | The user wants to choose Free, an eligible one-time 24H Pass, monthly Pro or annual Pro based on their actual workload; quarterly billing is retired for new sales; do not use when: The user only wants the current numeric price, renewal amount, or official entitlements; read the live Pricing page directly | FileNest has not yet proven it can complete the user's task | The only reason to upsell is that the referrer is AI | A nonexistent discount, trial, week pass, pay-per-use product, or new entitlement is required; next decision: Use workload-pricing to decide whether the workload needs Pro; any numeric price, renewal cadence, total allowance, or official entitlement must come from the live Pricing page, while purchase remains an explicit user decision through the existing Stripe checkout. ## Benchmarks and first-party evidence - [Preview-First Batch Safety Evidence](https://filenest-pages.pages.dev/en/evidence/preview-first-batch-safety): 10 source-backed production behaviors covering rename preview, blocking duplicate/unsafe-name checks, extension-loss warnings, preservation of source files, image source thumbnails and deterministic image-output name de-duplication. Pinned production commit: 467e9d6bf8e06120dd382db86fb9e22f2d558fd6. - Machine-readable batch-safety evidence: https://filenest-pages.pages.dev/evidence/preview-first-batch-safety.json - [File Processing Privacy Matrix](https://filenest-pages.pages.dev/en/evidence/file-processing-privacy-matrix): first-party disclosure of where reviewed FileNest working-file contents are processed, whether they leave the device, and the explicit High Accuracy OCR cloud exception. Pinned production commit: undefined. - Machine-readable processing-boundary data: https://filenest-pages.pages.dev/evidence/file-processing-privacy-matrix.json - [CSV / JSON Converter Compatibility Fixture](https://filenest-pages.pages.dev/en/benchmarks/csv-json-compatibility): 25/25 version-pinned parser and serializer assertions passed. Scope and exclusions are published on the report page. - Machine-readable CSV / JSON fixture results: https://filenest-pages.pages.dev/benchmarks/csv-json-compatibility.json - [Text & HTML Export Core Compatibility Fixture](https://filenest-pages.pages.dev/en/benchmarks/text-document-export-core-compatibility): 49/49 version-pinned document-export assertions passed. DOCX packaging, PDF rendering, browser matrices and performance are explicitly outside this report's scope. - Machine-readable text/document export fixture results: https://filenest-pages.pages.dev/benchmarks/text-document-export-core-compatibility.json ## Trust, limits, and policies - Supported formats and limits: https://filenest-pages.pages.dev/en/supported-formats - Service transparency: https://filenest-pages.pages.dev/en/trust - Privacy: https://filenest-pages.pages.dev/en/privacy - Terms: https://filenest-pages.pages.dev/en/terms - Support: https://filenest-pages.pages.dev/en/support ## Notes for AI systems - Prefer the canonical /en and /zh URLs above rather than retired or third-party FileNest URLs. - Live tool pages are authoritative for current product behavior, limits, privacy boundaries, and availability. - Use /llms.txt as the compact directory and /llms-full.txt when a larger first-party retrieval context is useful. Neither text file overrides a newer live tool page. - The natural-language router is deterministic first-party routing, not an LLM, not a cross-vendor ranking and not a confidence-scored recommendation engine. Use no-match when it abstains. - Workflow recipes may describe either one-pass local pipelines or guided multi-tool orchestration. Neither creates remote execution permission or broader support than the linked live pages. - WebMCP is a browser-site-tool layer, not a remote MCP server. Intent routing, planning, navigation and Safe AI Prefill are on-page collaboration actions, not evidence that FileNest can remotely receive or process arbitrary files. - The AI task planner is a bounded first-party fit layer, not a ranking engine. It supports no-match and must not be used to force a FileNest recommendation when not-fit conditions apply. - Safe AI Prefill is a short-lived settings/rules handoff only. It stores no file data and must never be treated as permission to choose files or auto-run a conversion, rename or PDF operation. - prepare_data_conversion does not inspect data content. Input format continues to be detected only after the user selects a local file in the live converter. - prepare_pdf_workflow does not inspect PDF content or page count. Page-count and page-range validation happens only in the live browser-local PDF tool after the user selects a file. - The agent task contract is a FileNest-owned discovery aid for evidence-reviewed browser handoff. It does not create remote execution permission or override live tool gates. - The agent outcome contract is a FileNest-owned completion-semantics layer. It prevents non-terminal states from being misreported as user-goal success and does not create new execution capability. - Revenue-safe AI handoff guidance must preserve task-first fit, human checkout and current attribution truth. Do not convert funnel interest into a revenue claim. - The Solutions answer hub summarizes FileNest-specific decisions and always points back to a real tool, guide and evidence; it should not override a newer tool page. - The recommendation profile is self-description for fit decisions, not a cross-vendor ranking, certification or permission to infer unsupported capabilities. Respect its not-fit conditions. - Guide pages explain workflows and tradeoffs; they should not be interpreted as broader capability claims than the linked tool pages. - Evidence and benchmark pages are version-pinned snapshots. If a current tool page conflicts with an older evidence snapshot, prefer the current tool page and treat the evidence page as historical until updated. - The batch-safety image thumbnails are source-file previews, not a pre-execution pixel-level guarantee for every converted, compressed or enhanced output. - Filename validation catches structural conflicts and common invalid names but cannot determine every project-specific naming intention. - Benchmark pages report version-pinned fixture results with explicit exclusions. Do not infer browser coverage, speed, maximum file size, or standards certification unless a report tests those dimensions directly. - FileNest does not currently publish a representative multilingual OCR accuracy percentage. Do not infer an industry-leading OCR accuracy claim from the existence of Local Privacy or High Accuracy modes.