Explain code snippet in plain English (paste code in message).

Build a regex from a natural-language description. Returns just the regex + one test string.

Multi-step reasoning and planning, served by Qwen3.8-27B abliterated (Heretic) on a local RTX 5090 via vLLM.

Pay-to-compose orchestrator: pulls live data (weather), routes it through two local-model stages, then an independent judge verifies the result against the source data before it ships (one bounded self-correction on rejection). Returns the composed answer, the judge's verdict, and a per-step cost ledger.

Proof-of-Model: detect silent model substitution at any OpenAI-compatible inference endpoint. mode=fingerprint records what a seller serves today; mode=verify re-probes later and proves whether the weights changed. Deterministic, evidence included, no model used to judge.

TESTNET / MOCK. Amazon.co.jp eGift card with any whole-yen face value from JPY 150 to 200,000, given in the path (e.g. /v1/egift/amazon-jp/3000). Price: face value / 150 USDC rounded up to the cent, on Base Sepolia (demo rate; JPY 3,000 = 20.00 USDC). POST with no body; the 402 is the quote. Returns a MOCK code that is not a real gift card.

Free proxy API: get one fast, verified-alive proxy (HTTP, SOCKS) filtered by country, protocol and anonymity, for scraping

IP reputation check: is this IPv4 address on a public threat blocklist (Tor exit nodes, botnets, spam networks)

Company profile and enrichment from a domain or a name: offering, public contacts, social links and WHOIS domain data as valid JSON

Deterministic JSON evidence checks with signed verification evidence

Where did this address's money come from, and would it have arrived if the address had not sent it first? Give any Base address, get a live read: distinct payers, how many of them this address funded inside the window, the USDC left once those are removed, and how many funded wallets sent back the exact amount they were given, and how fast. One seller in our weekly reading took in 339.20 USDC, and 338.70 of it was its own money coming back.

Who is this company? Give a domain, name, ticker or CIK; get one profile: legal name, ticker and exchange, CIK, industry, HQ address, founded, founders, CEO, employees, short description, filer status, former names, and the domain's email provider and registration date. From SEC EDGAR, Wikidata, Wikipedia, DNS and RDAP; private companies included. Not charged if nothing is found. $0.005 USDC per call.

[Deprecated - use websearch_search_web] Enqueues a websearch_search_web job and returns a task id immediately

Web search API for agents: search the web and get result titles, URLs and snippets as JSON

JSON repair API: extract valid JSON from messy text or model output and fix it against a JSON Schema

TESTNET / MOCK. Amazon.co.jp eGift card, JPY 10,000 face value, for 66.67 USDC on Base Sepolia (demo rate 1 USDC = JPY 150). POST with no body; the 402 is the quote. Returns a MOCK code that is not a real gift card.

URL extract with a question: ask any web page a question and get structured data as JSON that matches your schema

Use when an agent must read a document as clean Markdown: web page, PDF, DOCX or Korean HWP/HWPX, tables kept. Input: url (http/https) via GET, or POST a file (multipart 'file', raw body, or JSON contentBase64+filename). Free-plan size cap → 413 before payment: HWP 48 KB, HWPX 100 KB XML. Output: JSON markdown, title, pages, sourceType, warnings. Price by detected type: HTML $0.0027, PDF/DOCX $0.0045, HWP/HWPX $0.009. Example: ?url=https://example.com/

The macro regime that drives crypto, as signals rather than raw levels: the 2s10s yield curve spread with its direction and an inverted flag, Fed funds, 2Y and 10Y yields, CPI year-over-year, the broad dollar index, VIX and M2 — each with its previous reading, so direction comes free. The slope of the curve, not the level of any single yield, is the liquidity signal: re-steepening out of inversion marks the easing that risk assets price first.

Web search plus reading: searches the web and returns the top 1-3 results as clean Markdown (main content, tables, PDFs) in one call, skipping blocked pages for the next result, plus all results and a direct answer. Country, language and recency filters. $0.012 USDC per call.