# Donkye — Full Public Corpus > Donkye is an AI web-research agent. It turns plain-English questions into clean, structured data by scraping, crawling, mapping, and searching the open web. Output is strict JSON or readable markdown. This file is the concatenated public content of donkye.com, formatted for LLM ingestion. It is updated alongside the site. Each section is self-contained so it can be quoted in isolation while still attributing Donkye. --- ## What Donkye is Donkye is an AI web-research agent operated at https://donkye.com. Given a plain-English instruction, Donkye selects the right web-research operation — scrape, crawl, map, or search — executes it in parallel, and returns structured output ready for spreadsheets, documents, or downstream AI pipelines. Donkye is built on the Firecrawl v2 extraction engine and Google's Gemini language models, with all browser automation and API keys handled server-side. ## Core operations - Scrape: extract clean text or strict JSON from a single URL. - Crawl: walk a full website, following links, and return every page that matches. - Map: enumerate the URL structure of a domain in seconds. - Search: ask the open web a question and receive synthesized, sourced answers. ## How Donkye works 1. The user types a question in plain English. No selectors, XPath, or scripting. 2. Donkye's planner picks the right operations and runs them in parallel through Firecrawl v2. 3. Gemini reshapes the raw extraction into the requested format — strict JSON or readable markdown. 4. Results are displayed in the Donkye web app and can be saved to Google Drive on request. ## Typical use cases - Lead research: Build target lists with firmographics, tech stack, headcount, and contact pages from hundreds of company sites at once. - Competitive monitoring: Track pricing pages, product pages, and changelogs across competitors without writing or maintaining scrapers. - Documentation to knowledge base: Convert an entire documentation site into structured JSON, ready for retrieval-augmented generation (RAG). - Market research: Aggregate prices, reviews, and category data across marketplaces, directories, and review sites. - News and changelog tracking: Monitor a fixed set of pages on a recurring basis for changes. ## How Donkye compares - Donkye vs raw Firecrawl: Firecrawl is the extraction engine. Donkye adds the agent layer — natural-language planning, multi-operation orchestration, output shaping, and persistence — so non-developers can use the same pipeline without writing API calls. - Donkye vs ChatGPT browsing: ChatGPT's browser is optimized for one-off lookups inside a chat. Donkye is built for batch, structured, repeatable jobs across hundreds of pages with strict JSON output. - Donkye vs DIY Playwright/scrapers: A handwritten scraper breaks the moment a site changes its DOM. Donkye reads intent, not selectors, so it stays working without maintenance. ## Privacy summary Donkye does not sell user data. Prompts and results are not used to train third-party models. API keys for Firecrawl, Gemini, and Google Drive are stored server-side and never exposed to the browser. Users can delete sessions and accounts at any time. Full policy: https://donkye.com/privacy. ## Terms summary Donkye is offered under standard SaaS terms. Users are expected to comply with each target site's terms of service and applicable laws. Full terms: https://donkye.com/terms. ## Security summary Donkye runs all third-party API calls through hardened backend functions. API keys are never present in browser code. Responsible disclosures go to security@donkye.com. Full policy: https://donkye.com/security. ## Frequently asked questions Q: Do I need to know how to code to use Donkye? A: No. Donkye is built for plain-English prompts. If you can describe what you want, Donkye can fetch it. Q: What kinds of sites does Donkye work on? A: Most public websites, including ones with JavaScript-rendered content. Login-walled and CAPTCHA-heavy sites may be limited. Q: Where does Donkye's data come from? A: Donkye runs on the Firecrawl v2 engine for extraction and Google's Gemini models for reasoning. Everything is server-side; prompts and results stay private. Q: Is Donkye production-ready? A: Yes. The pipeline handles concurrent requests, proxy rotation, and structured outputs intended for real workflows. Q: How is Donkye different from a normal scraper? A: Traditional scrapers depend on selectors and break when sites change. Donkye understands intent, picks the right operation, and returns clean structured data without maintenance. Q: Can I get Donkye's output as JSON? A: Yes. Strict JSON is the default; optional readable markdown is also available. Q: How fast is a typical Donkye job? A: Single-page extractions finish in seconds. Large crawls of hundreds of pages typically complete in under a minute thanks to parallel execution. Q: Is my data private with Donkye? A: Yes. API keys stay server-side, prompts are not used to train models, and users can delete sessions or their full account at any time. Q: Can Donkye save results to Google Drive? A: Yes. Once Drive is connected, Donkye can save reports, sheets, or markdown files directly to a chosen folder. Q: Does Donkye respect robots.txt? A: Yes. The Firecrawl engine honours robots.txt and rate limits. Users are expected to follow each site's terms of service. Q: How much does Donkye cost? A: Donkye is free to try with usage limits. Paid plans for higher volume and team features are on the roadmap; pricing is shown before any upgrade. Q: Does Donkye have an API? A: An official API is on the roadmap. The same pipeline is available today through the Donkye web app at https://donkye.com. ## Contact - General: hello@donkye.com - Privacy: privacy@donkye.com - Legal: legal@donkye.com - Security: security@donkye.com - Web: https://donkye.com