Eligibility data
For AI agents: this page is also Markdown at /developers/eligibility-data.md. Every docs page is listed in /developers/llms.txt.
What the 233 eligibility rulings mean, how matching works, and how many link a source.
The API serves 233 expense rulings. Each has one of three values in eligibility:
| Value | Count | Meaning |
|---|---|---|
eligible | 160 | Qualifies as a medical expense under IRS rules for most people. |
partially | 46 | Qualifies only in some cases, often with a Letter of Medical Necessity or a diagnosis. See requirements and conditionalEvidenceTypes. |
ineligible | 27 | Does not qualify. Paying for it from an HSA is a taxable, possibly penalized, distribution. |
Sources
47 of 233 rulings carry a sourceBasis array: each entry names a source (usually an IRS publication or IRS FAQ), links it, says what part of the ruling it supports, and gives the date we last reviewed it. The other rulings are our summaries of IRS guidance without a linked source yet. We add sources over time; keep only rulings with a non-empty sourceBasis if you need cited answers.
How the check endpoint matches
exact: the query is the item's name, slug or a known alias.foundis true.strong: the query starts or contains the name, or shares its key words.foundis true. Queries under four letters count only as whole words (ivfyes,ptno).- Anything weaker never answers.
foundis false,matchis null, and up to three weak hits appear inalternativesfor a "did you mean".searchUrlpoints to the full search.
Phrases like "is", "HSA eligible" and "can I use my HSA for" are stripped before matching, so you can pass the user's question as typed.
Freshness
meta.updated is the date this API build was published, not a per-record review date. Each sourceBasis entry has its own reviewedDate. Contribution limits come from the IRS revenue procedure for each year.
Not tax advice. A ruling describes the general rule, not your plan's rules.