one way to call every listed tool.

Search the open web for ranked links (title, url, snippet, score). The default web search; use `read` to fetch a result's full text.

Fetch a URL and return clean markdown, no nav or clutter.

Search the web for ranked links (title, url, snippet). Use `read` to fetch a result's full page.

Scrape a URL to clean, LLM-ready markdown with JS rendering. Handles dynamic, script-heavy pages a plain reader misses.

Search the web and get ranked results (title, url, description). Fetch a result's full text with the read or firecrawl_scrape tool.

Current conditions for a place by name or lat/lon.

Structured brand profile (logos, colors, description, socials, industry, address) for a company by domain, email, name, or ticker.

Classify a company into 2022 NAICS industry codes (code, name, confidence) by domain or company name.

Classify a company into SIC industry codes (code, name, major group, confidence) by domain or company name. type: original_sic (default) or latest_sec.

Resolve a company or brand name to its domain — fuzzy search over logo.dev's brand index.

Ahrefs Domain Rating (DR) for any site, 0-100 on a logarithmic scale. Display it with the 'Domain Rating by Ahrefs' attribution the response carries.

The 100 strongest sites on the web by Ahrefs DR, with their rank.

Web, news, or research-paper search with domain include/exclude lists, a freshness window or date range, and country/language targeting. Pass intent (why you are searching) for better-ranked results.

Read up to 10 URLs in one call as markdown, html, or a structured tree. CSS selectors scope what is extracted, and ttl accepts a cached copy instead of a live fetch.

Independent scores, pricing, and measured speed for every tracked LLM: the Intelligence, Coding, and Agentic indices, input/output/cache price per 1M tokens, and median output tokens per second, time to first token, and end-to-end response time. 200 models per page.

Head-to-head arena rankings for one media modality: text-to-image, image-editing, text-to-video, image-to-video, either video-with-audio arena, text-to-speech, speech-to-speech, speech-to-text, or instrumental and with-vocals music.

Web results from Brave's own index, not a rented one: title, url, and snippet, with freshness windows down to the last 24 hours, country and language targeting, and search operators.

News articles with publisher, age, and thumbnail, filtered to the last day, week, month, year, or an exact date range.

Images from across the web with source page, dimensions, and thumbnail. Up to 200 per call.

Videos with duration, publisher, creator, and thumbnail, filterable by how recently they were posted.

Token volume per model per day across all of OpenRouter, the numbers behind its public rankings, with an 'other' row summing everything outside the top 50 so a share of total traffic is one request. Slice it by modality, by request context length, or by what the traffic was for: category and language_type read a sampled weekly dataset and cannot be mixed with the exact modality and context_bucket filters.

Which apps put the most tokens through OpenRouter, with request counts. sort 'trending' ranks by growth against the three preceding periods instead of raw volume, so it surfaces what is climbing rather than what is already large.

The last 7 days of OpenRouter traffic split by task: code generation, web search, summarization and the rest, each with its share of requests and of tokens, the top models within it, and a roll-up into Code, Data, Agent and General. Shares of sampled traffic, so no absolute volumes.

Median USD spend per session for the published agent harnesses, broken out by model and by how many turns the session ran. Filter by model alone to compare harnesses at a fixed model. Refreshed weekly.

Prompt a Groq-hosted model with the OpenAI chat shape, including tools, reasoning and structured output. Defaults to openai/gpt-oss-20b. Every chat field passes through except streaming; the response comes back whole on your own key.

Groq's beta OpenAI Responses API for text or image input, reasoning, function calls, browser search, code execution and remote MCP tools. Stateful previous_response_id is not supported. The call uses your own key and returns whole, not streamed.

Create one or more text embedding vectors with Groq. Defaults to nomic-embed-text-v1_5, accepts up to 2,048 input strings, and can return float arrays or base64-encoded vectors on your own key.

Transcribe a public audio or video URL with Whisper on Groq. Defaults to whisper-large-v3-turbo; use verbose_json with word or segment timestamp_granularities for timings. Groq fetches the URL directly.

Daily share of real Vercel AI Gateway traffic, by model, by lab, by app, or by inference provider, counted in requests, token volume, spend, and images or videos generated. models and labs give one row per entity per day back to 2025-10-01 and can be narrowed to a modality; apps and providers give a ranked top list with no day dimension. Shares only, never absolute volumes. Published by Vercel under CC BY 4.0.

The whole catalog across every creator: per-token pricing for input, output and cache reads and writes, tiered by context length where that applies, plus context window, max output tokens, capability tags, knowledge cutoff, supported parameters, and the zero-data-retention and no-training flags. Covers image, video, embedding, speech, transcription, realtime and reranking models, not only chat. Filter by type or creator; the whole list is large.

Pricing, context window, max output, capability tags, supported parameters, knowledge cutoff and retention flags for a single model, without pulling the whole catalog.

Every inference provider serving one model, each with P50 and P95 time-to-first-token and tokens per second measured on the last hour of real gateway traffic, uptime over 15 minutes, an hour and a day, and its own pricing, context length, quantization and supported parameters. The same model can run an order of magnitude apart between providers at near-identical list prices, so this is the tool that picks one.

Turn a product URL into a structured listing draft: name, tagline, descriptions, category, tags, pricing and platform type. No account needed.

The free-tier launch calendar with remaining capacity per week (20 slots each). Featured listings skip this queue.

List your own submissions with their status.

One of your submissions in full.

Indexable startup directories ranked by Ahrefs Domain Rating, each with its pricing and whether the link is dofollow. The shortlist for where to submit a product.

Search ScrollLaunch's directory of launched indie products by keyword, category or tag.

The ranked product leaderboard for an ISO week, ordered by upvotes. See what launched and how it did.

Full metadata for one launched product: tagline, description, categories, website and upvotes.

Write a post to your DEV account. Leave `published` false to create a draft and review it on dev.to before it goes live. Set canonical_url when the piece also lives on your own site.

Edit one of your posts, or flip a draft to published by passing published true.

Take one of your posts back to draft. The post is not deleted.

Your own posts, including drafts. Use state to narrow to published or unpublished.

Search the web with an optional synthesized answer (set include_answer). Good for quick factual lookups.

Search the web and get a written, synthesized answer (not just links). Use when you want an answer, not sources.

Run several web search queries in one call (pass search_queries). Use to batch multiple distinct queries.

Turn public URLs (including JS-heavy pages and PDFs) into markdown.

Kick off a deep web research run and get a run_id back. Poll it with task_result. processor: lite (fast/cheap), base (default), or core (deepest).

Fetch a deep-research run's result by run_id. Long-polls ~30s; if still running it returns { status: "running" } so you can call again.

Fetch the latest remote job listings.

Host and share a web page: publish a rendered markdown or interactive HTML page (report, dashboard, landing page, chart, slides, mini-app) to a shareable fetchbean URL a human opens in a browser. fetchbean handles the styling (Tailwind + a base stylesheet) and hosting; you send the content. Returns the link. For a completed local PDF, DOCX, PPTX, or XLSX, use the multipart file-artifact flow in the guide.

Revise an existing artifact in place (same URL). The prior version is kept for rollback.

Surgically edit an artifact with find/replace edits instead of resending the whole body — same URL, a version is snapshotted first. Best for small changes to a large artifact.

List your published web and file artifacts with durable ids, links, lifecycle state, engagement, and file metadata. Returns 100 at a time by default (500 max) alongside `total`, so page with limit + offset when total exceeds what came back.

Search iOS App Store apps by keyword and storefront country.

Fetch full App Store metadata for an iOS app by track ID.

Fetch App Store reviews with rating distribution for an iOS app.

Fetch stored App Store rating snapshots for an iOS app.

Find meetings by keyword, date, or participant; returns matching transcript ids and metadata (not full text). Use first, then the transcript tool.

Fetch a meeting transcript with its summary, keywords, and action items.

Get just the action items and outline from a meeting.

Confirm whose Fireflies account the connected key acts as.

Run a HogQL/SQL query over your PostHog events, persons, and analytics. The main PostHog read tool.

Your PostHog account, organization, and accessible projects (ids for the other tools).

List your feature flags (prefer this curated tool over the raw request escape hatch).

Get one feature flag by id.

The Linear account your key authenticates as.

List issues, filtered by team, state, assignee, or priority.

Get one issue by UUID or human identifier (e.g. ENG-123).

Free-text search across your Linear issues.

List the projects your key can see.

Get ONE customer by their app_user_id (the customer_id param), incl. active entitlements.

List/page customers; the optional search is exact-email only (no name or partial match).

A customer's subscriptions (status, store, access, renewal).

The Expo account/robot your token authenticates as, and its accounts.

List the apps (projects) under an account; gives the app_id the build/update tools need.

Get one app by full name (@owner/slug) or app_id.

Recent EAS builds for an app (status/platform/artifact url); filter by platform (ios/android) or status.

Any user's contribution calendar (the green graph): total + daily counts. Omit from/to for the last year.

A user's public profile (bio, company, location, follower/repo counts).

A user's public repositories, newest-pushed first (stars, language, description).

Search repositories, users, or issues (type = repository|user|issue).

List your uptime monitors with status, last check, and SSL/domain expiry.

Get one monitor by id.

List monitor groups.

List heartbeat (cron/job) monitors with their status.

List devices in the connected tailnet, including addresses, ownership, status, and client version.

Get one tailnet device by its stable device id.

Read the subnet and exit-node routes advertised and enabled for a device.

List users in the connected tailnet and their role and status.

Get one project (description, stars, default branch, visibility).

List/search projects (search=, membership=true for your own).

A user by username or numeric id.

Read a file from a project repo (decoded text). ref = branch/tag/sha.

Full-text search across pages and databases. filter='page'|'database' narrows the object type.

Get one page's properties + metadata by id (use read_page for the body content).

Read a whole page's content as markdown — recursively crawls the block tree and flattens it (capped for very large pages).

Get a database's schema (title, properties) by id.

List payments (filter by status, customer_id, etc.).

Get one payment by id.

List subscriptions (status, customer).

Get one subscription by id.

List the domains in your account (expiry, auto-renew, nameservers).

Details for one of your domains (expiry, auto-renew, nameservers, status).

List the DNS records for one of your domains.

Check whether a domain is available to register.

List your Supabase projects (id, ref, name, region, status).

Edge functions deployed to a project, with slug, status and version.

Preview branches on a project, with their status and parent.

Per-service health for a project — database, auth, storage, realtime.

List your zones (domains). Filter by name.

Details for one zone by id.

List a zone's DNS records (filter by type/name).

Every setting on a zone and its current value — SSL mode, minify, HTTP/3, security level and the rest.

List customers (filter by email, etc.).

Get one customer by id.

List payment intents.

Get one payment intent by id.

List channels (public/private the bot can see).

Every reply in one thread, given the parent message's ts. Channel history returns only parent messages, so this is the one that reads the actual conversation.

Search the workspace with Slack's own query syntax — in:#channel, from:@user, before/after a date. Needs a user token; a bot token cannot search.

User ids in a channel.

List the connected user's servers and show which are installed/default.

Get one connected server; omitted guild_id uses the default.

Add the connected user to a server where the fetchbean bot is installed.

Leave a server as the connected user. Server owners must transfer ownership first.

List the organizations your token can access.

Teams in the organization and the projects each owns.

People in the organization, with role and the teams they belong to.

Raw error events across every project, filterable with Sentry search syntax. Issues group events; this is the ungrouped stream underneath.

Confirm whose Trakt account the connected token acts as.

Your watchlist, optionally filtered by type (movies, shows, seasons, episodes).

What you've watched, newest first. Filter by type and/or a date range (start_at/end_at).

Your ratings (1-10), optionally filtered by type.

Search movies, TV, and people. type='movie'|'tv'|'person' narrows it (default searches all). Returns TMDB ids + poster_path.

Full movie detail by TMDB id: overview, runtime, genres, plus credits, videos, images, external_ids (imdb_id), and watch providers.

Full TV-show detail by TMDB id: overview, seasons, genres, plus credits, videos, images, external_ids, and watch providers.

Trending titles. media_type='all'(default)|'movie'|'tv'|'person', time_window='day'|'week'(default).

The 116 Indian cities PVR INOX operates in, with coordinates and cinema counts. Filter by name/state (query) or region (NORTH/SOUTH/EAST/WEST).

Cinemas in a city with distance and show counts. `film` keeps only cinemas showing that title. Distance is measured from lat/lng — pass your own, because a city's stored coordinates are not always its centre (Bengaluru's sit ~20 km out). full=true adds every film playing at each cinema, which is most of the response.

Films in a city — status='now_showing' (default), 'coming_soon' for the release calendar (dates, cast, director, synopsis, trailer), or 'all'. Filter by query, language, genre, or experience (imax, 4dx, luxe). The city's premium formats come back in `exp`; full=true adds its cinemas and the long-form format list, ~1.4MB that pvr.cinemas serves better.

Showtimes across a city or near a point, one block per cinema. date is YYYY-MM-DD (today if omitted; other dates cost one upstream call per cinema, so narrow them with `film`). Filter by film, language, experience (imax/4dx/luxe), time window (after/before, '18:00' or '6:00 PM'), available_only (drops Sold Out and already-started Lapsed shows), accessible_only. A whole city is covered by default; pass lat/lng or cinema_id to stay local. Each show carries an `encrypted` token for pvr.seats and pvrcinemas.com/seatlayout/<token>. full=true re-attaches each cinema's whole film list, which is ~85% of the bytes and already answered by the showtimes themselves.

The Conductor account the connected key belongs to: user id, email, organization, and how it authenticated.

Repositories you can create workspaces in. Start here — creating a workspace needs a projectId. Paginated with limit/offset.

One project by id, with its name and git remote.

Workspaces belonging to one project. Paginated with limit/offset.

Every subdomain on file for an apex domain, from Certificate Transparency, each with the date it was first seen. `apex` must be the registrable domain (example.com, not sub.example.com). Returns total alongside the page; limit defaults to 500, max 5000. This is a history of names that were ever certificated, so some no longer resolve, and anything never issued a certificate of its own is absent.

List users (filter by query, email_address, user_id, order_by).

Get one user by id.

List organizations (filter by query, order_by).

Get one organization by id or slug.

The Buffer account your key authenticates as, and its organizations (with organizationId).

List connected channels (social profiles) for an organization; gives the channel_id posts need.

List posts for an organization (Relay-paginated); filter by channel_ids or status.

Aggregate post analytics over a time window (ISO start/end), optionally per channel.

A profile with follower/following/post counts: the connected account by default, or any actor (handle or did).

The connected account's home timeline (the following feed), paginated with a cursor.

Posts authored by an actor (the connected account by default). filter e.g. posts_no_replies, posts_with_media.

The connected account's notifications (likes, reposts, replies, follows, mentions, quotes).

Get the Render user associated with the connected API key.

List workspaces the connected API key can access.

Retrieve one workspace by id.

List services, databases, workers, cron jobs, and static sites. Filter by ownerId, name, type, region, etc.

The connected cal.com account (id, email, username, timezone).

Your cal.com event types (id, title, slug, length). The eventTypeId here is what you book against.

List your bookings (filter by status or attendeeEmail; paginate with take/skip).

Open time slots for an event type between start and end (dates). Book one with create_booking.

Your Canny boards (id, name, post count). The boardID here is what you filter and post against.

Feedback posts: search, filter by board/status/tag, sort by score, trending, newest or statusChanged. Statuses: open, under review, planned, in progress, complete, closed.

One post in full by its id: details, status, score, author, tags.

Comments, filtered by post, board, author or company. Paginate with the returned cursor.

Turn text into speech in any of ElevenLabs' voices (voice_id from elevenlabs.voices; defaults to Rachel). Returns the audio as base64 plus an audio_url that serves the same bytes from fetchbean, so you can hand off a link instead of megabytes.

Describe a sound and get it as audio — footsteps on gravel, a UI chime. Leave duration_seconds off to let the model pick. Returns the audio as base64 plus an audio_url that serves the same bytes from fetchbean, so you can hand off a link instead of megabytes.

Transcribe speech from a URL with Scribe, with word timings. ElevenLabs fetches the URL itself, so the file must be reachable without auth. Set diarize true to label who spoke.

The voices on your account, including any you cloned. The voice_id here is what speak takes.

Text to speech across Indian languages (bn-IN, en-IN, gu-IN, hi-IN, kn-IN, ml-IN, mr-IN, od-IN, pa-IN, ta-IN, te-IN (translate also takes as, brx, doi, kok, ks, mai, mni, ne, sa, sat, sd, ur)), code-mixed text included. bulbul:v3 caps at 2500 characters. Sarvam answers with base64 in 'audios'; that is passed through untouched, with an audio_urls copy served from fetchbean.

Transcribe speech with Saaras, tuned for Indian languages and code-mixed speech. Give a URL under 25MB; fetchbean fetches it and uploads the bytes, since Sarvam takes only a file.

Translate between Indian languages and English (bn-IN, en-IN, gu-IN, hi-IN, kn-IN, ml-IN, mr-IN, od-IN, pa-IN, ta-IN, te-IN (translate also takes as, brx, doi, kok, ks, mai, mni, ne, sa, sat, sd, ur)). source_language_code defaults to auto. mode picks the register: formal, modern-colloquial, classic-colloquial or code-mixed.

Convert text between scripts without translating it — Hindi in Latin letters into Devanagari, or the reverse. Meaning and words stay; only the script changes.

Your support inbox: conversations filtered by status (open, resolved, pending, snoozed), inbox, team, labels or assignee_type (me, unassigned, all). Paginate with page.

One conversation with its contact, assignee, labels and status.

Advanced search: an array of {attribute_key, filter_operator, values, query_operator} clauses, for anything the plain conversations filters can't express.

How many conversations are mine, unassigned and all — the queue depth, without pulling the list.

Your Intercom inbox, newest first. Paginate with per_page + the starting_after cursor.

One conversation with its full part-by-part thread. Bodies come back as plaintext unless you ask otherwise.

Find conversations by state, assignee, tag, channel or time. query is Intercom's search shape: {field, operator, value} or {operator: 'AND', value: [...]}. Paginate with per_page + the starting_after cursor.

Users and leads in the workspace. Paginate with per_page + starting_after.

The Plain workspace your key is scoped to, and the permissions it holds.

List the customers in your workspace.

Get one customer by Plain id, email address, or your own external id.

Partial-match search across customer name, email, short name, and external id.

The Astronomy Picture of the Day with its explanation: today by default, a specific date, a date range, or a number of random days. Videos come back as links, and thumbs asks for a still.

Asteroids making a close approach to Earth in a date window (7 days max), each with its miss distance, relative velocity, estimated diameter, and whether NASA flags it as potentially hazardous.

The full record for a single asteroid by its JPL SPK-ID, including orbital data and every logged close approach.

Space weather events from NASA's DONKI database: FLR solar flares, GST geomagnetic storms, CME coronal mass ejections, SEP solar energetic particles, HSS high-speed streams, and more. Defaults to the plain-language NOTIFICATIONS feed.

Exchange rates for 201 currencies: today's by default, a specific day with date, or a series with from/to. Rates are blended from 84 central banks and reach back to 1948, and every day is covered including weekends. A range must name its quotes; use group week or month to downsample a long span.

Convert an amount from one currency to another, at today's rate or a past date's. Returns the converted result alongside the rate used and the date it applies to.

Every currency code available, with the providers covering each. include_legacy folds in currencies no longer in circulation, such as the pre-euro national ones.

The 84 central banks and official sources behind the rates, each with the currencies it covers, its rate type, publishing cadence, and how far its history goes.

Look up a US-listed company by ticker, CIK, or name: its CIK, every ticker and exchange it trades on, industry (SIC) code, EIN, fiscal year end, website, and business address. A name matching several companies comes back naming them.

A company's filings, newest first, each with a direct link to the filed document. Narrow with form (10-K, 10-Q, 8-K, 4, S-1) and since. Covers roughly the last 1000 filings; the response says when older ones exist beyond that.

Search the full text of filed documents, not just their titles. Quote a phrase to match it exactly. Each hit names the company and form and links straight to the document. Narrow by form type and date range.

One reported financial figure for one company, every period it appears, straight from XBRL — so revenue or net income comes with the filing and form it came from. Common tags: Revenues, NetIncomeLoss, Assets, Liabilities, StockholdersEquity, CashAndCashEquivalentsAtCarryingValue.

Search ~160M scholarly works by title, author, or topic, each with its DOI, authors, journal, year and citation count. Sort by is-referenced-by-count for the most cited. Matching is bibliographic rather than semantic, so a title works better than a plain-English question.

Resolve a DOI to its full record: title, authors, journal, publication date, publisher, licence and how many works cite it. Accepts a bare DOI or a doi.org link.

Earthquakes from the USGS catalog with magnitude, depth, coordinates, felt reports and tsunami flag, timestamps already in ISO rather than epoch milliseconds. Defaults to the last 30 days; narrow by magnitude, by date, or to a radius around a point. Says how many matched in total, not just what came back.

Everything a VIN encodes, from the manufacturer's own filing: make, model, year, trim, body class, engine size and cylinders, horsepower, transmission, airbag positions, assembly plant and GVWR class. Says whether the check digit validated, so a plausible-looking bad VIN is distinguishable from a clean decode. Inapplicable fields are dropped rather than returned as 'Not Applicable'.

Every make registered with NHTSA — 12,000-odd, counting one-off custom builders — or just those building one type of vehicle (car, truck, motorcycle, bus, trailer). NHTSA offers no search on this table, so `contains` filters the answer here.

The models a manufacturer sold, for one model year or across every year on file. The reference list the other vehicle tools expect a model name from.

Open safety recalls on a vehicle, with the campaign number, the component at fault, what goes wrong and what the remedy is. Carries NHTSA's two urgency flags — park it, park it outside — and whether the fix is over-the-air. Dates are converted from NHTSA's day-first format to ISO.

What fueleconomy.gov actually rates, one level at a time: every model year with no arguments, the makes sold in a year with `year`, and that make's models with both. The spellings the mileage lookup expects.

Official EPA figures for every engine and transmission combination of one vehicle: city, highway and combined MPG, estimated annual fuel cost, CO2 grams per mile, the EPA fuel-economy and greenhouse-gas scores, emissions standard and electric range. Covers everything sold in the US since 1984.

Average US pump prices per gallon by grade — regular, mid-grade, premium, diesel and E85 nationally — for today, yesterday, a week ago, a month ago and a year ago, so a change is visible without a second call. Pass a two-letter state code for that state plus each of its metro areas.

Today's average price for every US state and DC in one call, per grade — the table to rank, compare or find the cheapest and most expensive states from.

The weekly average US mortgage rate from Freddie Mac's Primary Mortgage Market Survey — the 30-year and 15-year fixed figures quoted in the news. Returns this week's rate on its own, or the whole weekly series between two dates: the 30-year runs back to April 1971, so a rate today can be put against any point in fifty years.

The monthly state of the US housing market from Realtor.com's open research data, for the whole country, a state, a metro, a county or a ZIP code: median listing price, price per square foot, active and new listing counts, median days on market, how many listings cut or raised their price, and pending sales — each with its month-over-month and year-over-year change. Sort by any column to rank the hottest or cheapest markets. Covers roughly 28,000 ZIP codes.

Every post office served by an Indian PIN code, with district, state, postal circle and division, and whether each one delivers or is non-delivery. One PIN usually covers several post offices, so this is what turns a six-digit code into a place you can name and validate an address against.

The PIN codes for a named post office, locality or neighbourhood in India, with its district, state and postal hierarchy. The other direction from a PIN lookup: for when an address has a place name and needs the code.

The bank, branch, address, city, district and state behind an Indian IFSC code, plus its MICR and SWIFT codes — and, the part that decides whether a payment goes through, which rails the branch is live on: NEFT, RTGS, IMPS and UPI. Validates the code's format before asking, so a typo comes back named rather than as a bare not-found.

Look up a packaged product by its barcode: name, brand, ingredients, allergens, and nutrition per 100g, plus the three headline scores, Nutri-Score A to E for nutrition, NOVA 1 to 4 for how processed it is, and Eco-Score for environmental impact. Four databases sit behind this: 'database' defaults to food and also takes beauty (cosmetics), petfood and product (everything else). A barcode in the wrong one comes back naming the right one.

Search packaged foods by name, brand, or category, each hit carrying the same nutrition and score fields as a barcode lookup. Food only: the beauty, petfood and product databases have no search index, so reach those by barcode.

Prices people have photographed off the shelf: the amount, the currency, the date and the named shop. Filter by barcode, by category tag, by a lat/lon and radius to see what a thing costs near a place, or by date to watch it move. Coverage is partial and leans European, about 135,000 of Open Food Facts' 4.5 million products carry any price at all, so a barcode with no prices is the common case rather than an error.

The shops behind the prices, each with its address, coordinates and how many prices and products it has on file. Names come from OpenStreetMap in the local language, so 'country' matches Deutschland rather than Germany.

Rank about 1,370 cities on what it costs to live there for a nomad, expat, local or family, rent and coworking, internet speed, the weather and air quality right now, and scores out of 5 for safety, cleanliness, fun, nightlife, walkability, healthcare, English, and how LGBT- and female-friendly it is, plus whether it has a remote work visa. Filter by region, country, budget, minimum internet, minimum safety and a temperature range. Ordered best first, so no filters gives the live top ranking.

The same curated set of costs, scores and live weather for a single city, by the slug a search returns, for example lisbon-portugal.

Where and when nomads are meeting in person over the next days: venue, local time and how many people are going. Filter by city slug or country. Attendance is a headcount only, never who is going or who organises.

All 1,374 cities in rank order, each with latitude and longitude, nomad and life scores, cost of living, internet speed, temperature, air quality, population and photos. This is the list behind the map, so it carries coordinates and rank that a search hit does not, and fewer of the detailed scores. Narrow it with region or country, since the whole list is 1.5MB.

Search Hacker News stories and comments, with points, comment count and a link to the discussion. Narrow by type (story, comment, show_hn, ask_hn, front_page), by min_points, or since a date. sort date gives newest first instead of most relevant.

One story or comment by id, with its full nested comment tree — the way to read what a thread actually said.

Search arXiv preprints across physics, maths, CS, biology and economics. A plain phrase searches everything; prefix it to narrow — ti: title, au: author, abs: abstract, cat: category (cs.LG, hep-th). Each result carries its abstract, authors, categories and a direct PDF link. sort date gives newest first.

Fetch a preprint by its arXiv id (1706.03762, or 1706.03762v5 for a specific version), with abstract, authors, categories, DOI, journal reference and PDF link. Accepts several ids separated by commas, or a full arxiv.org URL.

Read a preprint's actual text rather than its abstract — introduction, method, results and all — by id or arxiv.org URL. A typical paper comes back around 38,000 characters against 1,100 for the abstract. A paper arXiv has not rendered falls back to the abstract and says so in `source`. max_chars caps the response.

Weather for a place by name or coordinates: current conditions and a daily forecast up to 16 days, with conditions written out rather than left as a WMO code. Times are local to the place. Override the variables with current, daily or hourly if you want something specific.

Recorded weather for a past date range, anywhere, back to 1940. Answers what the weather actually was on a day, or how a month compares to the same month in other years.

Air quality for a place: PM2.5, PM10, ozone, carbon monoxide, and both the US and European AQI, which use different scales and will not agree. Current plus a short forecast.

Daily river discharge from the GloFAS global flood model, up to 210 days ahead. Discharge rising far above its usual range is the flood signal; the number alone means little without that comparison.

Satellite scenes covering a place or bounding box, with the date, cloud cover and direct links to the imagery. Give a place name and it searches a box around it, or pass an explicit bbox. max_cloud matters more than it looks: over somewhere monsoonal, most scenes are unusable, so filtering can legitimately return nothing. Collection ids differ per catalogue — see stac.collections.

The collections available in a catalogue — Microsoft Planetary Computer carries well over a hundred, USGS LandsatLook the Landsat archive, AWS Earth Search open Sentinel and Landsat. Ask here before searching, because a collection id from one catalogue returns nothing on another rather than erroring.

A single scene by its id, with every asset link. On Planetary Computer the links are signed for you, since unsigned ones are refused; those signatures expire about an hour after they are issued.

Internally displaced people by country and district, from IOM's displacement tracking. Answers at the national total unless you name an admin1 or admin2, and returns the most recent reporting rounds first — the raw dataset is unordered, so an unsorted read hands back whichever decade it happens to start on. Distinct from humdata.refugees, which counts people who crossed a border.

Refugee and asylum-seeker counts from UNHCR, as flows: 'origin' is the country people fled, 'asylum' is where they are now. Give either to see one side of the movement, or both for a single corridor. At least one is required, because the unfiltered dataset is every corridor on earth.

People in need, targeted and reached, broken down by sector — health, protection, food security, shelter and the rest — from the humanitarian needs overviews. Filter by sector to answer 'how many need clean water in Yemen' rather than the headline figure.

What a humanitarian appeal asked for, what it actually received, and the percentage covered, from OCHA's financial tracking. Appeals run back to the 1990s and are returned newest first, so the current one leads.

Merchandise trade from UN Comtrade: what one country exported or imported in a year, in US dollars, by partner and commodity. Leave partner out for the world total, or name one for a single corridor. flow is X for exports, M for imports. Commodity is an HS code — find it with un.commodities. One year per call, because the open endpoint allows one period per request, and it is rate limited to about a request a second.

Look up the Harmonised System code for a product, which is what un.trade filters on. Codes nest by length: 09 is coffee, tea and spices, 0901 is coffee, 090111 is green coffee beans, so a shorter code is the broader category. Broadest matches are returned first.

All 17 goals with the 169 targets under them, or one goal with its own. This is the map: it tells you which goal a question belongs to before you go looking for the series that measures it.

Search the 700-odd statistical series behind the SDG indicators by keyword or goal, and get the series code un.sdg_data needs. Search here first: an indicator like 1.1.1 is a concept, while a series like SI_POV_DAY1 is the thing with numbers attached.

Who the token belongs to: email, team, verification status, and the droplet and volume limits on the account. The quickest check that a connection is live and which team it reaches.

Current month-to-date usage, account balance and the last invoice amount. Month-to-date is what has accrued since the last bill, not a forecast of the next one.

Past invoices with their amounts and periods. Pair with digitalocean.balance for what is accruing now.

Every droplet on the account with its status, size, region, IP addresses and tags. Filter by tag_name to scope to one fleet. Pages are 20 by default and cap at 200.

Site-wide counters: tests started, tests completed and total seconds spent typing across everyone who has ever used it. Started and completed differ a lot, because an abandoned test still counts as started.

The distribution of results in 10 wpm buckets, so you can see where a speed sits against the field. Defaults to english time 60, the standard test. The counts are results, not people — a fast typist appears once per test they finished.

The top typists for one test. A board only exists per language, mode and length — english time 60 is the default and the one most people mean — so there is no overall ranking to ask for. limit has a floor of 10.

The best results of the current day for one test, which resets daily and is a far more reachable board than all-time. days_before 1 gets yesterday's final standings.

Your Clarity project's numbers for the last 1 to 3 days: traffic, engagement time, scroll depth, popular pages, and the frustration signals Clarity is actually useful for — rage clicks, dead clicks, excessive scroll, quickbacks and script errors. Break them down by up to three of Browser, Device, Country/Region, OS, Source, Medium, Campaign, Channel or URL. Two hard limits worth planning around: only ten requests per project per day, and no history beyond three days — this API cannot answer what happened last week. It returns aggregates only; session recordings and heatmaps are not available through it.

Generate an image from a text prompt with gpt-image-2.5, which renders text inside images far better than diffusion models do. Flare (default) is the fast everyday model; sunburst is slower and more precise. Quality runs low to max, size is any WIDTHxHEIGHT up to 4K, and background: transparent gives a cutout. Returns base64 rather than a URL, because OpenAI offers no URL option: higher quality and larger sizes mean a multi-megabyte response, so leave quality at low unless detail matters. To change an image you already have, use /images/edits. Runs on your own OpenAI key if you connect one, otherwise on fetchbean's.

Change an image you already have, or make a new one from up to 16 reference images, with gpt-image-2.5: 'swap the sign to CLOSED', 'put this product on a marble counter', 'the same character in a rainy street'. Inputs are URLs, data: URIs or base64. A mask limits the edit to its transparent pixels. Sunburst is OpenAI's pick when edit precision matters; flare (default) is faster. Returns base64 like /images/generations, and the input images add to the price on fetchbean's key. Runs on your own OpenAI key if you connect one, otherwise on fetchbean's.

Make an image from a prompt, or edit one you already have by passing it alongside the prompt — the same call does both, so 'put this product on a marble counter' is one request with the product photo attached. Input images are base64 or a data: URI, not URLs. Defaults to gemini-3.1-flash-image; gemini-3-pro-image handles complex scenes and legible text better, gemini-3.1-flash-lite-image is cheapest and 1K only. aspect_ratio (16:9, 1:1, 9:16) and image_size (1K, 2K, 4K) shape the output. Returns base64, because Google offers no URL option, so a 4K image is a multi-megabyte response. Runs on your own Google AI key.

Your Calendly user: name, email, timezone, scheduling URL, and the canonical URIs for you and your organization. Most other endpoints identify people by those URIs rather than by email, though the tools here resolve yours for you.

The meeting types people can book — the 30-minute intro, the demo — with their duration, scheduling URL and whether each is active. Defaults to yours; pass a user URI for someone else in the organization. This is where the event_type a scheduling link or availability lookup needs comes from.

Scheduled meetings with their times, status and who is on them, soonest first. Defaults to yours — pass organization to see the whole team's instead. start_after and start_before take a plain date or a full timestamp; a bare date covers the whole day. status 'canceled' finds what fell through.

Full detail for a single meeting: location, guests, host memberships, and the cancellation if it was called off. Takes the uuid or the full event URI.

Conversations in the inbox, newest activity first, each with the visitor, the last message and its state. Filter to unresolved to see what still needs answering, or search with query. Pages are numbered from 1 and hold about 25.

A single conversation's metadata: the visitor, assigned operator, state, segments and the last message. Use crisp.messages for the transcript itself.

The transcript, with who sent each message and when. Crisp returns the most recent batch; pass before with the earliest timestamp you have to walk further back rather than a page number.

Sends a text message into a conversation as an operator — the visitor sees it and is notified, so this is a real reply, not a draft. from 'user' instead writes as the visitor themselves, which puts words in their mouth in their own transcript; leave it alone unless that is genuinely what you want.

Boards you can reach, most recently modified first, with their name, owner, team and view link. Search names with query. Pages by offset and caps at 50 — the item listing pages by cursor instead, which is the one inconsistency in this API worth remembering.

Full detail for a board: owner, team, sharing policy, permissions and links. Use miro.items for what is actually on it.

The items on a board with their content, position and size. Filter by type — sticky_note, text, shape, card, image, document, frame, embed — to read just the notes off a workshop board. Pages by cursor, not offset: pass the cursor from the previous response to continue.

A single item on a board, with its content, geometry, style and parent frame.

Live positions from the OpenSky receiver network: callsign, country, latitude and longitude, altitude, speed and heading for everything currently flying over a bounding box. Give lamin/lomin/lamax/lomax, or an icao24 address to follow one aircraft. This is the cheap half of OpenSky — a box under 25 square degrees costs 1 credit against a daily budget, the whole world costs 4 — and the answer says what it charged and what is left. Coverage is volunteer receivers, so it is dense over Europe and thin over oceans and much of Africa.

Departures from an airport in a time window, with the aircraft, callsign and estimated arrival airport for each. airport is a 4-letter ICAO code (EDDF, KJFK, VIDP), not IATA. begin and end are unix seconds, defaulting to the last hour and capped at 24. Worth knowing before you loop over airports: the historical endpoints cost about 30 credits a call against a separate budget from live positions, where a whole day is 400 credits unconnected and 4,000 connected, so this is roughly thirteen calls a day on the free tier. An empty result means nothing was seen in that window, which is normal at a quiet airport rather than an error.

Arrivals at an airport in a time window, with the aircraft and its estimated departure airport. Same shape and price as departures: 4-letter ICAO code, unix-second window defaulting to the last hour and capped at 24, about 30 credits a call.

Every flight a single aircraft made in a window, by its icao24 address — the 6 hex digits that identify the airframe itself, not the flight number, so it follows the physical plane across the day rather than one scheduled service. Window defaults to the last hour and is capped at 24; about 30 credits a call.

The cheapest recent fares on a route, with the dates, airline and number of stops. Origin and destination are 3-letter IATA codes — use travelpayouts.places to find one. Leave destination out to see the cheapest fares from a city to anywhere. Read these as observed prices rather than live availability: they are the fares other travellers actually found, cached, so a figure here tells you what a route usually costs, not what you can book this second.

A price for each departure day in a month on one route, which is the tool for 'when is it cheapest to go'. month takes YYYY-MM. Same caveat as prices: a cheap-looking day is a fare somebody found recently, not a seat held for you.

The lowest fare found for each date on a route, for picking the month rather than the day. Pass by 'return' to organise it around the return date instead of the departure.

Popular destinations from one city ranked by how cheap the fares have been — the 'anywhere' search, for when the question is where to go rather than what a known route costs.

Find a model on Replicate by name or description — the way in when you don't know the exact owner/name.

Public models on Replicate, newest or most-run first.

One model, including its latest version and that version's openapi_schema — the authoritative list of input fields to pass to replicate.run.

Every published version of a model; the id is what replicate.run takes as version.

Live-check the connected Google account, its exact OAuth scopes, and which gws services those scopes can call.

Search the live Google Discovery catalog for methods supported by a gws service and see scope availability.

Get Google's live Discovery schema, parameters, request body, response, scopes, and media support for one method.

Call any gws-backed Google REST method directly from its Discovery id, with refresh, scope checks, pagination, and media transfer.

Send or schedule an email. Resend validates the sender, recipients, content, templates, attachments, and account quota.

List sent emails and their latest delivery event.

Retrieve a sent email, its content, tags, schedule, and latest delivery event.

Update a scheduled email using Resend's current request fields.

Test the connected API key and return its Loops team.

Retrieve one audience segment.

List audience segments for campaign targeting.

Create an audience segment using Loops' current request fields.

List events on a calendar (default 'primary'). Optional timeMin/timeMax (RFC3339), q, maxResults.

Get one event by id from a calendar (default 'primary').

List the calendars on the user's calendar list.

Create an event on a calendar (default 'primary'). event needs start + end.

Read a document's full content (body, styles, revision id) by document id.

Create a new empty document with a title. add content afterwards with doc batch update.

Apply edits to a document. requests is an array of docs edit ops (insertText, replaceAllText, etc.).

Read a value range from a spreadsheet (A1 notation, e.g. 'Sheet1!A1:C10').

Send an email as the connected account. to is required; subject, body (plain text), html, cc, bcc optional.

Run a GA4 report for a property. Pass numeric property_id, dimensions + metrics (name arrays), and a date range.

List the GA4 accounts and properties you can access, to find a numeric property_id.

Query Search Console traffic for a verified site over a date range. dimensions can be query, page, country, device, or date.

List the sites you have verified in Search Console.

Inspect a URL's index status in Search Console. Needs the inspection_url and its owning verified site_url.

List the sites verified in your Bing Webmaster Tools account.

Add a site to your Bing Webmaster Tools account. Adding is not verifying — pair it with bing_webmaster.verify_site, and until that passes the stats and submission tools have nothing to read.

Ask Bing to check a site's ownership proof now. Returns true once it passes; false means the meta tag, XML file or DNS record it expects is not in place yet. Read the codes to publish from bing_webmaster.sites.

Remove a site from your Bing Webmaster Tools account. This drops its history and verification; re-adding it means verifying ownership again.

Keyword search across Strale's capabilities and bundled solutions. Each match carries its slug, price in EUR cents, and category — feed the slug to execute_capability.

Describe the job in a sentence and get one recommended capability or solution with the reason it matched and its test-pass record, plus alternatives and how many others matched. Use when a keyword search does not name the thing you want.

One capability's input schema, output example, price, data source, and freshness. Read this to learn what `inputs` execute_capability expects for a slug.

Run any Strale capability or bundled solution by slug (or pass `task` and let Strale pick one), charged to your prepaid EUR wallet. `max_price_cents` is your spend cap for the call; `dry_run` reports what would run and what it would cost without spending. A capability over 10s answers with status pending and a transaction_id — poll it with transaction_detail.