Kcalbin turns public records into things you can actually use — layoff registers, bankruptcy dockets, federal spending, patent claims — and sells them at the smallest unit that makes sense. That might be a single $1 query, a single $395 candidate, or a monthly feed. This page is the directory: what each product is, who it is for, what it costs, and where it stops.
One convention runs through all of it. We say what a product does not cover on the product page, before you pay, rather than after. Where coverage is two states, this page says two states. Where something is a demo, it says demo.
WARN Act layoff data for employment attorneys
Two separate products read the same public source — state WARN registers — and answer two different questions.
WARN Intelligence scores each public WARN filing by county displacement risk and ranks it against every other live filing, so the question it answers is which of these filings is worth my attention first. A single scored lead is $149 one-time. Every CRITICAL and HIGH filing as it is scored is $299/month. There is a browsable set of scored leads at /warn/leads — 60 are listed there right now.
WARN Short-Notice Intelligence answers a narrower and more specific question: where does the state's own register already show fewer than the statutory 60 days between the notice date and the layoff effective date, at a site with 50 or more affected workers. The arithmetic is done for you and the candidate arrives identified. A single candidate is $395; the ongoing feed is $795/month. The live candidate list is at /warn/short-notice — 120 candidates are listed there right now.
The honest limits, which are also on the product pages: coverage is California and New York, not national. And a short interval is a screening signal, not a legal conclusion — the statute carries exceptions that can lawfully shorten the period (the faltering-company exception, unforeseeable business circumstances, and natural disaster, 29 U.S.C. § 2102(b)), and WARN coverage itself turns on employer headcount and the proportion of the workforce affected. Confirming that is the buyer's work. Kcalbin is not a law firm and does not give legal advice.
County-level displacement data, delivered over MCP
Chandler Displacement Intelligence scores workforce displacement at the county level, combining WARN filings, bankruptcy court data and BLS forecasts. The distinguishing part is delivery: it arrives inside Claude and other AI assistants over MCP, so you ask a question in the assistant you already use rather than downloading a file and building a spreadsheet.
Three tiers: $299, $999 and $2,999 per month. There is a free masked preview on the product page, so you can see the shape of the data before paying for it.
If you are searching for a county unemployment or layoff data source and what you actually want is a queryable signal rather than a raw BLS download, this is the one to look at.
Twenty data agents, from $1 a query
The agent catalogue holds 20 single-purpose data agents. Each wraps a real, named public or licensed source and answers one kind of question well. A single query is $1.00 to $2.00 by card, and 30 days of access is $12 to $30 depending on the agent. No wallet, no subscription minimum, no account required to buy one answer.
The current catalogue, with its source:
- GovSpend — federal and municipal government spending lookup (USAspending plus NYC, Chicago, Cook County and Delaware open data)
- Compute & GPU Pricing — Azure retail GPU and compute prices by region
- Flight Status & Airport Delays — FAA nasstatus and OpenSky
- Earthquake Alerts — USGS FDSN
- Currency Exchange Rates — ECB via Frankfurter
- Asia Equity Data — Twelve Data, commercially licensed
- US Company Registry & Insolvency — SEC EDGAR and CourtListener
- Product Recalls & Safety — CPSC, NHTSA and openFDA
- Vehicle Recall & Safety by VIN — NHTSA vPIC, recalls and NCAP
- Semiconductors & Export Controls — trade.gov CSL, UN Comtrade, Federal Register
- International Trade & Tariffs — UN Comtrade
- Elections & Campaign Finance — FEC
- Nonprofit 990 Intelligence — IRS 990 via ProPublica
- Climate & Wildfire Risk — NOAA and NIFC
- Local Crime Statistics — FBI CDE
- School & District Data — Urban Institute / NCES
- Restaurant Health Inspections — NYC and Chicago open data
- Transit Real-Time Arrivals — BART public API
- Sports Data — TheSportsDB
- ZIP Validation & Location — Zippopotam.us
Each is also callable agent-to-agent on the Virtuals Agent Commerce Protocol, so another autonomous agent with a funded wallet can buy an answer without a human in the loop. If you want the difference between that and MCP spelled out, we wrote a piece on what each protocol actually does.
Patent infringement evidence, either side
Patent Sentinel Evidence Service runs an adversarial analysis: prosecution and defense are argued independently, and then a judge returns an honest SUPPORTED / NOT SUPPORTED / INCONCLUSIVE verdict with cited sources. It serves either side — assertion or defense — and a verdict reflects the openable record, not adjudicated infringement.
It has no list price, because the shape of use varies too much for one. Access comes with a subscription: say the volume you expect and whether you are working assertion-side or defense-side, and a key comes back with the price for that shape of use. There is an open health endpoint so a buyer's agent can confirm the service is up before subscribing.
Custom AI data agent builds
If you have a data source and want it sold as an agent, the build service does the whole path: data integration, query logic and offering definition, ACP registration with outside-in verification, direct Stripe checkout for human buyers, and deployment as a monitored service.
- $6,000 flat, per agent — the full build
- $3,500 — each additional agent in the same engagement, priced down because the second one genuinely costs less once the infrastructure is standing
- $1,500 — registration and sale path only, if you already have the API and just need it registered, verified and wired to checkout
- $400 per agent per month for managed hosting and monitoring, or $1,500/month for up to ten agents. Cancel anytime.
No revenue share and no percentage of what your agent earns — the same policy as every other Kcalbin product. It runs on the same stack as the twenty agents above, which is the only real argument for it: the thing being sold is already in production here.
AUD Watch — in development, not for sale
On 15 September 2026 a fixed admission date replaces Duration of Status for F-1 and J-1 populations (DHS final rule 2026-14439). AUD Watch tracks admit-until dates against that rule.
Being straight about where it is: it is in development and not available to purchase. There is a working demo on the page, but every person shown in it is invented and it is labelled as simulated throughout. The date computation, spreadsheet ingestion and per-institution data separation are built and tested. Institutional sign-in is not built. Automated email alerting is not built. No real institution's data has been loaded anywhere. Until the alerting exists it is a dashboard you look at, not a system that tells you something has changed.
It is on this page because it is real work in progress, not because it is buyable. If you run an international student office and the rule change is on your desk, the demo is worth five minutes and the questions we would ask you are worth more to us than a sale right now.
Which one do you want?
- You are an employment attorney sourcing cases → WARN Intelligence for scored leads, or Short-Notice Intelligence for the sub-60-day subset.
- You want county-level economic and displacement signals inside your AI assistant → Chandler Displacement Intelligence.
- You need one specific fact from a government dataset, once → a data agent, from $1.
- You are working a patent matter on either side → Patent Sentinel.
- You have a data source and want it sold as an agent → the build service.
Everything above is priced and described as it stands on 17 August 2026. Where a product's coverage is narrower than its category suggests — WARN being two states, the agents each answering one narrow question — that is stated on the product page too. Questions, or a use case that does not fit any of these, go to contact.