US unemployment from BLS, monthly, newest first — national (CPS, "US") or by US state (LAUS). Pass area ("US" or a 2-letter state) and optionally measure = rate (default), unemployed, employed, or laborforce. Returns seasonally-adjusted values per month for a trailing window (rate in percent; counts in thousands). Public-domain, live-wrap. Fresh local labor-market context agents can't recite.

Occupations related to a given O*NET occupation (career-adjacent roles), by SOC / O*NET-SOC code. Returns ranked related occupations with code + title. CC-BY (O*NET). For career-pathing, "adjacent roles", and transferable-skills reasoning.

FDA 510(k) premarket clearances — the record of medical devices cleared for US marketing by demonstrating substantial equivalence to a predicate device. Search by device name, applicant (manufacturer), or FDA product code, and get back each clearance with its K-number, device name, applicant, decision date and decision description, clearance type (traditional/special/abbreviated), product code, advisory committee, and date received — newest decision first. Free, public-domain US government data. Authoritative regulatory-status data for medical-device diligence; point-in-time public records, not regulatory advice.

FDA drug approval history from Drugs@FDA — the official record of FDA-approved drug applications. Give a drug name (brand or generic), an FDA application number (e.g. NDA021436), or a sponsor name, and get back each application with its products (brand name, active ingredients + strengths, dosage form, route, marketing status, therapeutic-equivalence code, reference-drug flag) and its full submission history (original approvals, supplements, status + status date, review priority). Free, public-domain US government data. The authoritative "is this drug FDA-approved, by whom, when, and in what formulations" record — point-in-time public records, not medical advice.

Occupational employment and wages from the BLS OEWS survey, by SOC occupation code, nationally or for a US state. Pass soc (e.g. 15-1252 Software Developers) and optionally a 2-letter state. Returns total employment, hourly mean wage, and annual mean + 10th/25th/median/75th/90th-percentile wages for the latest survey year. Public-domain, live-wrap. Authoritative ground-truth wages an agent must not guess — for comp benchmarking, job-offer evaluation, and labor-cost modeling.

Generate a single combined AI-powered similarity key from a company/organization name plus a street address. Records with the same combined key likely share both a similar company name and address. Ideal for business-location deduplication across vendor, supplier, and account datasets.

SEC filings index for a company.

Search FCC Electronic Comment Filing System (ECFS) filings for a proceeding/docket (e.g. 17-108 net neutrality, 11-42 Lifeline). Optionally filter by filer name. Returns the newest filings with the submission id, filer, type, lead bureau, received/disseminated dates, and document count. Public-domain federal data. Track FCC regulatory dockets, comments, and ex-parte filings an LLM can't recite fresh.

AbuseIPDB subnet (CIDR) check — which IPs inside a network block have been reported for abuse. Pass network as a CIDR (e.g. 118.25.0.0/24; AbuseIPDB supports up to /16 for IPv4, /112 for IPv6). Returns the block metadata (network/netmask/min-max addresses, possible hosts, address-space description) plus reportedAddress: each flagged IP with numReports, abuseConfidenceScore, mostRecentReport, and country. Optional maxAgeInDays (1-365, default 30) and limit. Use to vet a hosting range, score a customer's netblock, or sweep your own allocation.

AbuseIPDB abuse report for a single IP — the crowd-sourced abuse-confidence score (0-100) the fail2ban / SSH-scanner / web-attack ecosystem reports into. Pass ip (IPv4 or IPv6). Returns abuseConfidenceScore, totalReports, numDistinctUsers, lastReportedAt, usageType (e.g. Data Center / Residential), ISP, domain, hostnames, isTor, isWhitelisted, and country. Set verbose=true for the individual report records (categories + reporter country). Optional maxAgeInDays (1-365, default 90) bounds the reporting window. Live crowd data an LLM cannot know — SOC alert triage, login-abuse blocking, and firewall decisions.

Resolves an IP address or domain to its Autonomous System Number (ASN), network owner, country, and hosting/cloud classification — agents get structured network ownership context plus a trust score to assess infrastructure risk.

Generate a single combined AI-powered similarity key from a company/organization name plus an individual's full name. Records with the same combined key likely share both a similar company and a similar person. Ideal for contact-level deduplication within CRM, marketing, and sales datasets.

Fetch public metadata for any GitHub repository — stars, forks, open issues, language breakdown, license, contributors, last commit date, topics, and an activity/health score — so agents can evaluate dependencies, research projects, and assess open source health without the GitHub API complexity. Pass repo as ?repo=owner/repo (GET) or {"repo":"owner/repo"} (POST).

Compare two individual/person names and receive a match score from 0-100 indicating how similar they are. Handles name order, nicknames, and abbreviations.

WHOIS domain lookup via RDAP — registrar, creation/expiry/updated dates, nameservers, and status flags from the authoritative registry, plus an actionability score so agents can assess domain trustworthiness, ownership, or availability risk.

Fetch a URL and detect the full technology stack from HTTP response headers, HTML meta tags, cookies, and script patterns — returns CMS, framework, CDN, server, analytics, and security tool detections so agents can profile any web target without a browser.

Reference registration/renewal/transfer pricing from a curated table of ~21 common TLDs — Mode A: price one TLD (typical register/renew/transfer cost + renewal-markup %); Mode B: price a name across all covered TLDs, sorted cheapest-first with cheapest/cheapest_premium picks. Reference prices, not live per-registrar quotes — for budgeting domain acquisitions and spotting predatory renewal markups; TLDs outside the table return pricing:null with an unknown_tld finding.

Parse any cron expression into a human-readable explanation, validate its syntax, compute the next 5 scheduled run times, and detect common patterns (daily, weekly, monthly, hourly) — so agents can understand, validate, and work with cron schedules without implementing their own cron parser.