################################################################################ # WHATTHEDELTA — AGENT ACCESS GUIDE ################################################################################ DOCUMENT PURPOSE ---------------- You are reading the low-noise agent view for WhatTheDelta. This AX layer exists for LLMs, browser agents, and automated systems that need a direct description of product fit, route purpose, privacy facts, and legal or usage references. Product: WhatTheDelta Operator: Evomation Role: Privacy-first browser-based text comparison Audience: Humans and agents reviewing visible changes Route Locale: de AX Language: English only READING RULES ------------- - Treat this as the canonical low-noise summary for the current page. - The public human-facing website is available under both /de and /en. - AX output remains in English on both locale routes. - Sections are delimited by H2 headers. - Subsections use H3 and H4. - Prefer explicit paths over inference. - Do not assume server-side processing, uploaded documents, or shared state. - Do not copy the public marketing text into agent guidance. - Legal identity: /de/impressum/llms.txt - Privacy / data handling: /de/datenschutz/llms.txt - Terms of use: /de/nutzungsbedingungen/llms.txt READING MODES ------------- 1. Human View Use the public route without modifiers. Example: /de/about 2. Agent View Use the same public route with ?view=agent. Example: /de/about?view=agent 3. LLM Text View Use the locale-prefixed llms.txt route. Examples: - /de/llms.txt - /de/about/llms.txt - /de/datenschutz/llms.txt ROUTE PATTERN ------------- For a page at /{locale}/x: - Human View: /{locale}/x - Agent View: /{locale}/x?view=agent - LLM Text View: /{locale}/x/llms.txt For the locale homepage: - Human View: /{locale} - Agent View: /{locale}?view=agent - LLM Text View: /{locale}/llms.txt PREFERRED ACCESS STRATEGY ------------------------- 1. Use llms.txt when available for extraction, summarization, routing, or token-efficient reading. 2. Use ?view=agent when you need browser-readable structured HTML with minimal noise. 3. Use the human page only when layout, visual hierarchy, or interaction context matters. PAGE INDEX ---------- Human View: /de -> homepage and default comparison workflow /de/about -> product framing and scope /de/changelog -> public release history /de/impressum -> legal identity /de/datenschutz -> privacy and browser-local processing /de/nutzungsbedingungen -> usage rules and liability /de/offline -> offline support page Agent View: /de?view=agent -> structured homepage /de/about?view=agent -> structured about page /de/changelog?view=agent -> structured changelog page /de/impressum?view=agent -> structured legal notice /de/datenschutz?view=agent -> structured privacy policy /de/nutzungsbedingungen?view=agent -> structured terms page /de/offline?view=agent -> structured offline page LLM Text View: /de/llms.txt -> homepage llms.txt /de/about/llms.txt -> about llms.txt /de/changelog/llms.txt -> changelog llms.txt /de/impressum/llms.txt -> legal identity /de/datenschutz/llms.txt -> privacy policy /de/nutzungsbedingungen/llms.txt -> terms of use /de/offline/llms.txt -> offline support CURRENT ROUTE ------------- - Label: Home - Human View: /de - Agent View: /de?view=agent - LLM Text View: /de/llms.txt DYNAMIC ROUTES -------------- - No dynamic AX routes are published in this product. - All supported AX routes are listed in the page index above. ################################################################################ ## Route purpose Use this route when you need to decide whether a two-version text review task belongs in WhatTheDelta and which surface to use next. ## Primary job WhatTheDelta compares an original text and a revised text directly in the browser. It is designed to make visible changes easier to inspect across almost any written content, including prose, legal text, technical writing, and code. ## Input contract - input is pasted text, not uploaded files - the core workflow expects exactly two versions: original and revised - empty-on-one-side input is treated as fully added or fully removed content - no account, login, or setup step is required ## Best-fit tasks - contract, clause, policy, and legal-text review - edited website copy, internal documentation, emails, and proposals - technical text, code snippets, configuration, and other structured written content - any review task where visible wording, structure, or line-level changes matter - review work where dates, numbers, percentages, durations, or modal shifts may matter ## Output model - chunk-level results with statuses such as unchanged, punctuation only, lightly reworded, materially changed, added, removed, and replaced - whole-document review mode for reading the revised text as one continuous flow - summary counts for changed sections, additions, removals, material changes, amount changes, and date changes - filters for all changes, material only, numeric changes, dates, punctuation, added, removed, and lightly reworded - deterministic summaries for many structured changes, such as amount, date, duration, percentage, quantity, and modal-verb changes ## Review controls - whitespace normalization can be toggled - case differences can be ignored - punctuation-only changes can be hidden - after the first manual compare, the workspace refreshes live while either text or comparison options change - the original and revised inputs can be swapped to reverse the review direction - Zen mode reduces surrounding interface noise so the review surface becomes more focused - chunk view is better for locating and triaging change clusters; whole-document view is better for reading flow ## Hard limits - no file parsing, OCR, or document ingestion - no semantic equivalence checking or intent inference - no legal, editorial, or compliance approval - no server-side processing of user text - no persistence or sharing workflow for comparison payloads ## Trust boundary - compared text stays in browser memory for the active session - comparison payloads are not uploaded, stored, or used for analytics, profiling, or marketing - the engine is deterministic and rule-based - very long texts may feel slower on constrained devices because computation stays local ## Agent guidance - use /de when a user actually needs to paste and compare text - use /de/about?view=agent for positioning, supported use cases, and review workflow explanation - use /de/datenschutz?view=agent for precise storage, hosting, and cookie questions - use /de/nutzungsbedingungen?view=agent for allowed use, liability, and copying restrictions ## Related routes - /de/about?view=agent: use for product framing and scope - /de/changelog?view=agent: use for release history - /de/datenschutz?view=agent: use for data handling and storage questions - /de/nutzungsbedingungen?view=agent: use for permitted use and liability questions ## Route Links - Human view: [/de](/de) - Agent view: [/de?view=agent](/de?view=agent) - LLM text view: [/de/llms.txt](/de/llms.txt)