Live USD price of any ERC-20 token, read from DEX pools at call time. Accepts a ticker (CAKE, WETH, ARB) or any contract address. Returns the price, the pair it was priced through, the USD liquidity backing that quote and a confidence rating. Refuses to return a price backed by a dust pool rather than reporting an unreliable number.

Market depth and TVL for a trading pair, broken down per venue: how deep the pools are, which venue holds the most liquidity, and each venue share of total depth.

Which DEX is cheapest to buy on and which pays most to sell into, with the cross-venue spread in bps and a fee-adjusted arbitrage spread. Smart order routing data.

Price impact for a SPECIFIC trade size — what a trade will actually get, which a spot price cannot tell you. Simulates the swap against live reserves on every venue. Price impact and the pool fee are reported separately. Pool-level only: excludes gas, MEV and multi-hop routing.

Pre-trade depth check for a pair: classifies the market DEEP, MODERATE, SHALLOW or VERY_THIN from live pool TVL, counts routable venues, and flags single-venue markets and wide spreads. Liquidity depth analysis only — NOT a contract audit and NOT a honeypot check.

Convert a street address, city or place name to latitude/longitude coordinates — worldwide, via OpenStreetMap. Returns coordinates, display name, structured address parts and a confidence score. No match returns a clear 404, never a guessed coordinate.

Convert latitude/longitude coordinates to the nearest street address and place name — worldwide, via OpenStreetMap. Coordinates in the ocean or unmapped return 404, never a fabricated address.

Cryptographically secure randomness for agents that are deterministic or sandboxed and cannot generate their own: uniform integers in [min, max] (rejection-sampled, no modulo bias) or raw random bytes as hex and base64. For nonces, IDs, sampling and shuffling.

Fetch a public article or PDF URL and return clean Markdown plus title, byline, site name and word count. HTML is extracted with Firefox reader-mode rules; PDFs return their text layer with page count. Image-only PDFs and client-rendered app shells return typed errors (no_text_layer, not_extractable) instead of empty output passed off as the article.

Web search — a free-text query returns ranked organic results, each with title, real destination URL, display URL and snippet. Sponsored rows are excluded. Up to 25 results; count is a maximum, not a guarantee, because a row whose destination cannot be resolved is dropped rather than guessed at. Results are not fetched or verified — pair with url_to_markdown to read any result.

Current weather and up to a 7-day forecast for any coordinates worldwide: temperature, feels-like, humidity, precipitation, wind speed/gusts/direction now, plus daily highs, lows and precipitation probability. Model forecast, not a station reading — the response says so. Use the geocode tool first if you have a place name rather than coordinates.

What people actually type into a search box for a topic — live search autocomplete, expanded across question modifiers (how/what/why/is/can/does) and comparison modifiers (vs/or), then split into suggestions, questions and comparisons. Rows where the engine dropped your term and answered something else are filtered out and counted. A demand signal for content and keyword research, not a ranking — search volume is not published upstream and is not invented.

Public and bank holidays for 100+ countries, and the business-day question behind them. Pass a date and it answers directly whether that date is a business day and what the next and previous ones are, counting weekends and holidays and walking across year boundaries. Without a date it returns the whole year, including which regions observe a non-national holiday. Regional closures, half-days and market trading calendars are not covered.

Read any public RSS, Atom or RDF feed as clean JSON: feed title, link and description, then items with title, real destination link, author, categories and the published date both verbatim and normalised to ISO 8601. Each summary is returned twice — the feed HTML as published, and a plain-text rendering safe to put straight into a prompt. An unparseable date is null, never an invented timestamp. Pair with url_to_markdown to read a linked article.

Locate a public IPv4 or IPv6 address: country, region, city, postcode, coordinates, timezone, ASN, org and ISP. Datacentre/VPN/proxy/Tor detection is returned FIRST because it decides whether the location means anything — and when the upstream does not supply that detection the answer says unknown, never clean. An IP locates a network, not a person: the coordinates are a registered-range centroid, so never present them as where someone is.

Look up a company in the GLEIF Legal Entity Identifier golden copy BY NAME, not just by identifier — knowing the LEI already is the hard part. Returns the LEI, registered legal name, previous names, legal form, jurisdiction, legal and headquarters addresses and the registration record. Lapsed, retired and annulled entities come back flagged rather than filtered out: a hidden record and no record are indistinguishable to the caller.

The US Treasury par yield curve for any published business day — every constant-maturity rate from 1 month to 30 years — plus the spreads and the inversion flag, because the real question is whether the curve is inverted rather than what fourteen numbers are. 2s10s, 3m10y and 5s30s are computed for you. A tenor Treasury did not publish that day is null, never zero. Nominal CMT rates: not real yields and not zero-coupon spot rates.

How much this session has spent on paid calls, and the caps in force. Free, local, and makes the cost of continuing visible before it is incurred.

Raw pool reserves for a pair on every indexed venue: the two token balances, the pool fee in bps, the implied price from those reserves and the pool TVL, all read at one block height which is returned with the answer. This is the underlying data the price, depth and slippage tools are computed from — use it when you want to do your own maths rather than take ours.

Scan a chain for cross-venue arbitrage right now: pairs whose price differs enough between DEXes to be worth trading, ranked by GROSS USD AT THE OPTIMAL TRADE SIZE — not by raw spread, because a wide spread on a tiny pool is not an opportunity. Returns an empty list when there is nothing, which is a real answer. Excludes gas, MEV and execution risk.

Live gas prices across BNB Chain, Polygon, Arbitrum, Base, Avalanche and Optimism, priced in USD and ranked cheapest-first. Returns gas price in gwei, base fee, and what a transfer, an ERC-20 transfer and a swap actually cost in dollars on each chain. Gwei is NOT comparable across chains because the gas token differs in price, so USD is the only ranking that says where a transaction really costs least. For bridging, routing and execution timing.

Supported chains, their indexed tokens and venues. Free, and the right first call if you are unsure which chain or ticker to use.

PAID ($0.01 USDC/call, no free tier). Real-time cross-DEX price & spread for BSC: per-venue prices across PancakeSwap v2, PancakeSwap v3 (all fee tiers), Biswap and ApeSwap in one call, plus best buy/sell venue, gross arbitrage spread (bps + USD), optimal trade size, liquidity and block number. Params: pair=SYM/SYM (e.g. WBNB/USDC), optional fee=v3 tier. Needs a funded wallet in DEX_WALLET_KEY or EVM_PRIVATE_KEY; without one it returns payment instructions and spends nothing. Cheaper alternatives in this same server, free for 25 calls/day and no wallet: "get_liquidity" for per-venue depth and "find_arbitrage" for a cross-venue spread scan.