AI assistants

Use Nodal in your AI assistant

Connect Nodal to an assistant that supports MCP connectors, such as ChatGPT, Claude, or Meta’s Muse. Then ask whether a product may qualify for HSA or FSA funds, and keep track of what you checked across the Nodal app, the Chrome extension and your assistant.

Nodal’s answers are general guidance based on IRS rules for the item, not your specific plan, and not tax, legal, or medical advice. Your benefits administrator makes the final reimbursement decision.

Connect Nodal

  1. In your assistant’s settings, add a custom connector (also called a remote MCP server) with this URL:
    https://api.nodalhealth.ai/mcp
  2. Your assistant opens Nodal’s sign-in page. Sign in with your Nodal account, or create one with Apple or your email.
  3. On the consent screen, choose Full access (check, list, save and dismiss items) or Read only (check, explain and list, but never save or change anything).

Then try: “Is the Omron Platinum blood pressure monitor HSA eligible?”, “What did I check in Nodal this week?”, “Save it to my Nodal list.”, or “Why does the massage gun need a letter from my doctor?”

Disconnect at any time at api.nodalhealth.ai/oauth/connections (sign in with your Nodal account), or by deleting your data in the Nodal app. The assistant loses access within a minute.

What an assistant can and cannot see

A connected assistant can seeIt never sees
The product details it sent, and the storeReceipts or receipt images
Nodal’s result, its reasons and the IRS sourceClaims or claim status
Items you checked or saved in Nodal on any deviceLetters of medical necessity
Where an item was added and whenAmounts you paid
Whether an item is checked, saved or dismissedWho an item is for, or answers to health questions

Assistants send product details only: a product link, a barcode number, or a product name with brand and store. Nodal removes anything that looks like health information before checking, refuses names that carry contact details, and never stores the rest of what the assistant typed. A check you don’t save is kept for 24 hours so the assistant can explain or save it. Nodal does not sell this data or use it for advertising. What Nodal returns to the assistant is covered by that company’s own terms. Full details are in our Privacy Policy.

Tools

The connector has five tools. Read tools change nothing. Write tools change only the user’s own Nodal item list and can be undone. No tool is a sensitive write: nothing deletes data on request, makes a purchase, moves money, submits a claim, or contacts anyone.

Tool Label Scope Side effects
check_eligibilityReadeligibility:checkNone. Saves nothing.
explain_verdictReadeligibility:check or items:readNone
list_my_itemsReaditems:readNone
save_itemWriteitems:writeAdds the item to the user’s list, or brings back a dismissed item. Repeating it changes nothing.
dismiss_itemWriteitems:writeHides the item. save_item brings it back for 30 days; after that it is deleted.

Every tool also carries the standard MCP annotations (readOnlyHint, destructiveHint, idempotentHint, openWorldHint). Every result comes back as plain text and as structured content, so assistants that show no app cards read just as well.

check_eligibility

Checks whether a product may qualify for HSA or FSA funds. Send one of a product link, a barcode, or a product name. When a link can’t be read and a name was also sent, Nodal checks the name.

Returns check_id (valid for 24 hours), status, label, summary, product (name and store), requirements, category, source (the IRS rule and link), question (when Nodal needs a product detail), what_to_keep, saved, can_continue_in_app (always false today) and disclaimer.

explain_verdict

Explains a result and what proof to keep for a claim. Input: check_id or item_id. Returns status, label, product, up to five reasons, requirements, what_to_keep, source, question and disclaimer.

list_my_items

Lists the user’s items from every surface, newest first. Inputs: state (any of checked, saved, dismissed; default checked and saved), limit (1 to 50, default 20) and cursor. Returns items and, when there are more, next_cursor. Each item has item_id, name, merchant_label, status, label, state, added_from and updated_at.

save_item and dismiss_item

save_item takes a check_id from check_eligibility, or an item_id to bring back a dismissed item. dismiss_item takes an item_id. Both return the updated item.

Results

statusLabelMeaning
eligibleEligibleA qualifying medical expense under IRS rules.
likely_eligibleLikely eligibleProbably qualifies. Confirm with the benefits administrator before paying with HSA or FSA funds.
lmn_requiredLMN requiredQualifies only with a Letter of Medical Necessity from the user’s own provider.
needs_more_infoNeeds more infoNodal can’t tell from the product details. Not a no.
not_eligibleNot eligibleGenerally not a qualifying medical expense under IRS rules.

requirements can include letter_of_medical_necessity, prescription and medical_use. Items are in one of three states: checked (seen by the extension, app or an assistant), saved or dismissed. Every check and explanation ends with the same advisory line: the plan administrator makes the final decision.

Authentication

Errors and limits

A tool that can’t finish returns isError with one plain sentence the assistant can pass on, saying what to change:

Each connection can make 120 MCP requests a minute and refresh its token 30 times a minute. Over the limit, /mcp answers 429 with Retry-After: 60. Each account also has a daily limit on eligibility checks.

Support and security

Last updated: 2026-10-30.