PDF to text from uploaded bytes. POST the raw PDF as the body (content-type application/pdf); multipart 'file' or JSON {"base64"} also work. Real PDF only (starts with %PDF), max 20 MB, with a text layer. Optional ?format=text|markdown. Bad files get HTTP 400 with a hint and are not charged. Sample: https://agents.crowdspark.me/samples/sample.pdf

Question answering over your own text: ask a question about material you supply and get an answer grounded only in that text (Gemini). If the text does not contain the answer it says so instead of guessing. Use it for summaries, Q&A and checks over documents, emails and pages. Input: text, question; optional language. Fixed JSON fields: llm.extract.

Structured data extraction / text to JSON: pull the fields you name out of any text into a JSON object with exactly those keys, null where the text does not state a value (Gemini, no invented values). Use it to turn emails, invoices, pages or documents into clean records. Input: text, fields[]. From a live web page: research.page_facts.

News search API: current news on any topic in any language from the GDELT Project, which reads the world's press every 15 minutes. Titles matched in their original script, newest first, with publisher, language and link; filter to one language; a US stock ticker searches the company's name. Window up to 30 days (default 72 hours). Input: query and/or symbol; optional language, limit, since_hours.

Flight status and airport delays, live: by airport code, FAA ground stops and delay programs with average and max delay (US), live departure and arrival boards with gates and status (Norway), current METAR and TAF weather (any ICAO airport) and aircraft overhead; by callsign, the aircraft's live position, altitude and speed. Input: airport (IATA or ICAO) or callsign; optional flight, direction, hours_ahead.

HTTP fetch / GET any URL: exactly what the server sent, status code, headers and body, plus the final URL after redirects. Every status is a result, so a 404 or 500 from the target is delivered, not failed. Use it for API checks, uptime probes and raw HTML. Input: url. Readable text from a rendered page: extract.page.

Shipping cost and delivery check for any Shopify store: the variant goes into a real cart session and the store quotes its shipping options to your destination. Returns options with prices and days, eligibility (or the store's reason, or the address fields still needed), subtotal and total before tax. No order is placed. Input: url, ship_to{country, province, postal_code}; optional variant, quantity.

Web search API / SERP results: raw web and news results from Brave's independent index in any language and country, with title, URL, description, age and extra snippets (headline, source and age for news). Filter by country, language and freshness (day, week, month, year). Results, not an answer: pick what to read next with extract.page. Input: query; optional country, language, count, freshness, news.

Text to speech (TTS): convert text to natural spoken MP3 audio with Google Cloud Text-to-Speech, returned as base64. Use it for voice replies, narration, audio versions of text and accessibility. Input: text; optional voice name.

Real estate data for any US address: FEMA flood and 17 other hazard risk ratings with expected annual loss, median home value, rent, income, ownership and vacancy (Census), price trend (FHFA), fair market rents (HUD), nearby public schools (NCES), walkability (EPA) and Superfund sites nearby, plus tract, county, ZIP and districts, all from federal sources in one call. Input: address, or lat and lng.

Twitter / X sentiment and volume for any topic, as numbers, live: post volume per hour or day over the last 1-7 days, and over a sample of recent posts the engagement totals and rate (likes, reposts, replies, quotes, impressions), language mix, top hashtags and link/media share. Aggregates only: no post text or authors. Input: query (X operators allowed); optional days, granularity, sample, lang.

Travel time with live traffic between two places, from the Google Routes API at call time: duration with traffic, duration without, the delay, distance, route description, warnings and advisories. Places as addresses, names or 'lat,lng'. Use it for ETAs, delivery windows and trip planning. Input: origin, destination; optional travel_mode, departure_time, traffic.

Deep web research with citations: search the live web for your question, read the top sources in full, and get a written answer with every claim tied to a numbered source. Use it when an agent needs a sourced answer, not a list of links. Input: question; optional max_sources. Quick answer with links: search.web.

BigQuery SQL: run a read-only SQL query, including against Google's public datasets (Wikipedia, GitHub, blockchains, weather, census and more), and get columns and rows back as JSON with the bytes processed. Every query is dry-run first and refused above the scan ceiling, so cost is bounded before it runs. Input: sql; optional max_scan_gib. Ask in plain English instead: data.question.

Dataset file to JSON rows: a CSV, TSV, XLSX or JSON link from any portal, an e-Stat (Japan) download or a data.go.kr (Korea) dataset page. Encoding detected (Korean, Japanese, Chinese, UTF and Western code pages), header row found, numbers typed, XLSX sheets selectable, large files capped and disclosed. Use it after opendata.search or on any data file URL. Input: url; optional max_rows, sheet, encoding, delimiter.

Weather forecast anywhere on Earth: current conditions, the next 24 hours and a seven-day forecast (temperature, precipitation, wind, humidity, cloud, pressure, condition) from MET Norway's forecast model, plus active US National Weather Service alerts for US locations. Input: location as a place name, US street address or airport code, or lat and lng.

Structured facts from a web page / scrape to JSON: name the fields you need (pricing, plans, contact, product names, anything) and get exactly those keys back as JSON from the live page rendered in a real browser, null where the page does not say. Use it for price monitoring, lead data and catalog extraction. Input: url, fields[].

Summarize and analyze a web page: the URL is rendered in a real browser, its content extracted, and a written brief answering your question returned, grounded in that page only and saying so when the page does not answer. Use it for competitor pages, docs, articles and listings. Input: url; optional question. Fixed JSON fields: research.page_facts. Needs a web search: research.web.

Website change monitoring, step one: fetch a page and store it as the baseline to compare against later. Use it to watch pricing pages, terms, job boards or competitors. Input: url. Then call monitor.check to learn whether and how it changed.

Run Python code in a sandbox: execute Python in Google's hosted sandbox (not on this service's machines) and get the code, its printed output and the outcome back. Use it for calculations, data transforms and checks an agent should not do in its head. Input: code. Questions over BigQuery data: data.query.