The standard way to get county-level economic data is to download it. BLS gives you a file, Census gives you a file, the state WARN agency gives you a file, and you write the same joining and cleaning code every organisation writes independently.
That workflow assumes the consumer is a person with a spreadsheet. When the consumer is an agent answering a question — *"which counties in this state have both a recent WARN filing and rising unemployment"* — the file format is the wrong shape entirely. The agent does not want a CSV; it wants an answer to a query it composed a second ago.
What MCP changes
Model Context Protocol lets a data source present itself as something an agent can query directly, rather than as a download an agent has to be taught to parse. The practical difference is that the joining logic lives with the data rather than being reimplemented by every consumer.
Chandler exposes county displacement data this way: 3,221 counties, each with real BLS county unemployment, WARN filings resolved to the county, and court signals — queryable rather than downloadable.
What it is genuinely good for
Screening. Narrowing 3,221 counties to the handful worth a human look, using more than one signal at once. That is a query, and it is tedious as a spreadsheet exercise.
Context on a specific county. You have a county and want to know how it sits against the rest — unemployment, recent filings, where it ranks.
Wiring into something else. If you already have an agent doing workforce or regional analysis, this is a source it can call rather than a dataset you maintain.
What it is not good for
Forecasting. The score is a relative ranking across counties on current inputs. It is not a prediction, and reading it as one will mislead you.
Anything where the middle of the distribution matters. The percentile is meaningful at the tails. Counties clustered around the median are separated by differences too small to carry weight.
Fine-grained sub-county work. County is the resolution. There is no tract or ZIP breakdown.
The inputs, named
BLS Local Area Unemployment Statistics for the county rate. State WARN filings from the agencies that publish them. Federal court signals. Those are the sources; where a figure in the output does not trace to one of them, it is not in the output — several derived fields were removed from this dataset for exactly that reason.
From $299.