The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Medicare Pricing listing page.
What Medicare pays — physician fee schedule, lab rates, and per-day unit
limits. The companion to medicare-coverage, which answers whether something is
covered but not what it is worth.
Part of Pipeworx — an MCP gateway connecting AI agents to 1683+ live data sources.
| Tool | Answers |
|---|---|
medicare_physician_payment | "How much does Medicare pay for 99213 in Los Angeles?" — RVUs and dollars, locality-adjusted. |
medicare_lab_rate | "What does a comprehensive metabolic panel reimburse?" — national CLFS rate. |
medicare_lab_test_lookup | "What does Medicare pay for Signatera, and which MolDX LCD applies?" — brand name → PLA code, current CLFS rate, and the MolDX LCDs/articles for the test's category. |
medicare_dmepos_rate | "How much does Medicare pay for a CPAP machine (E0601) in Texas?" — state-priced, rural vs non-rural. |
medicare_code_units_limit | "How many units of this code will they pay for in one day?" — NCCI MUE. |
check_code_pair | "Can I bill 00142 and 64474 together?" — NCCI PTP edit + modifier indicator for a code pair. |
explain_ncci_edit | "What is bundled into (or bundles into) this code?" — every recorded PTP edit for one code. |
medicare_pricing_coverage | Which vintages are loaded, including the PTP baseline vintage. |
None. No key, no account.
All from CMS, all downloadable files with no API:
medicare_lab_test_lookup exists because callers name a test by its brand
("Signatera"), not its code ("0340U"), and nothing else in the catalog maps one
to the other (fleet #2426).
Source, and why it is the only one. CMS's list of Advanced Diagnostic Laboratory Tests (https://www.cms.gov/files/document/advanced-diagnostic-laboratory-tests-under-medicare-clfs.pdf) gives code, lab, proprietary test name, approval date and initial-period payment for 18 tests. It is a US federal work (17 USC 105). The sources that were checked and rejected:
medicare-coverage handles it).cptcode field is
empty for them.That made coverage ADLT-only, and fleet #2458 measured the cost: Oncotype DX
(81519, a Category I MAAA code, not an ADLT) routed here 5/5 and came back
found: false. Resolution now runs through three sources in order:
src/adlt.ts, below).src/brand-codes.ts) -- non-ADLT branded tests whose
proprietary name is tied to a code in CPT Appendix O and which each have a
brand-named or category MolDX billing article on the CMS MCD (Oncotype DX
breast/DCIS/colon/prostate, Prosigna, MammaPrint, EndoPredict, Breast
Cancer Index, Cologuard, Prolaris, Decipher, Afirma, ConfirmMDx, 4Kscore).
Only brand -> code -> lab FACTS are kept, never AMA descriptor text. Every
code was checked to carry a CLFS rate on 2026-09-26. To add a row: confirm
the pair in a public source, confirm medicare_lab_rate has a rate, and set
coverage_topic to a phrase from a live MolDX article title.A test found in none returns found: false, reason: brand_not_found with
sources_checked naming all three, and a pointer to medicare_lab_rate.
Copy, not proxy. The list is a two-page PDF with no API, and the gateway
has no PDF parser, so the rows are baked into src/adlt.ts. The PDF's "Test
Descriptor" column is AMA PLA text and is not copied. Every call sends a
HEAD to the PDF and compares Last-Modified with the snapshot's. If CMS has
republished, the response carries snapshot_current: false. To refresh:
download the PDF, extract its text (pypdf works), update the rows in
src/adlt.ts, and bump ADLT_DATA_AS_OF and SNAPSHOT_LAST_MODIFIED.
Coverage. Each test has a coverage_topic, a phrase that is ours and not
CMS's. It matches the title of the MolDX LCD/article family for the test's
category (for Signatera, "Minimal Residual Disease"). The tool reads the live,
keyless CMS Coverage API reports and returns the active LCDs and billing
articles with that title. Retired ones come back only with
include_retired: true. This matters because contractors re-issue MolDX
policies under new numbers: Noridian's L38816 and A58456 were retired
2026-02-05 and replaced by L38814 and A58454. Whether a specific test is listed
is answered by the billing article's code table, not by the title match.
Fee schedules change annually and NCCI/CLFS quarterly. Quoting a 2026 procedure
at 2024 rates is a wrong number with billing consequences, so every response
states the year (and quarter where quarterly) it used, and sets
year_was_defaulted when the caller did not name one. Asking for a year that is
not loaded returns year_not_loaded rather than silently answering with a
different year's rates.
Medicare pays by geographic locality. California alone has nine. The same office visit (99213, 2026) pays:
| Locality | Non-facility |
|---|---|
| Bakersfield, CA | $99.52 |
| Alabama | $87.79 |
A 13% swing a state-shaped answer would hide. medicare_physician_payment
resolves the locality explicitly, states what it resolved to, and lists other
localities that matched so a caller can pick a different one.
Payment is computed with the real formula and the arithmetic is returned so it can be audited:
Facility and non-facility are both given — a procedure pays less in a hospital because the facility bills its own fee separately.
check_code_pair and explain_ncci_edit now cover the FULL NCCI
procedure-to-procedure (PTP) edit baseline, not a rolling change window
(fleet #1293, 2026-09-07). CMS distributes the complete PTP edit table
behind an AMA CPT licence click-through (/license/ama?file=... on
cms.gov). Bruce authorized accepting that licence on the company's behalf via
the Needs Bruce dashboard (2026-09-07); the licence text was read in full
before any data shipped — see "NCCI PTP licence" below for the quoted terms
and the reasoning for why this pack's serving shape stays inside them.
Loaded: the 2026 Quarter 3 baseline (v322r0, effective 2026-07-01 — the
newest vintage actually in effect; CMS had already posted Q4 2026 at ingest
time, but that isn't effective until 2026-10-01 and using it would have
answered today's queries with tomorrow's rules). The 8 baseline files
(practitioner + hospital × 4 code-range splits each) hold 4,498,118 raw
historical-episode rows — the same (column1, column2) pair can have several
add/delete/re-add episodes over the years — collapsed to 4,109,749 distinct
pairs by keeping each pair's open-ended (currently active) episode, or else
its most recently closed one. The prior 2025q1..2026q3 quarterly change
files are re-applied ON TOP of the baseline, in that order, so a pair the
quarterly files recorded as deleted cannot be resurrected by an older baseline
row.
A NOT-FOUND answer now means CMS has never recorded a PTP edit for that
pair — not merely "no recorded change in this window", which was the
scope limit before this fleet task. medicare_code_units_limit remains the
separate units edit (one code, not a pair) and was never affected by this
limit.
The CMS licence click-through (same boilerplate CMS uses across its whole AMA CPT-derived file family — the physician fee schedule, MUE, and PTP downloads all carry it) reads, in relevant part:
CPT codes, descriptions and other data only are copyright 1995-2025 American Medical Association. All rights reserved... You, your employees and agents are authorized to use CPT only as contained in the following authorized materials of Centers for Medicare and Medicaid Services (CMS) internally within your organization within the United States for the sole use by yourself, employees and agents. Use is limited to use in Medicare, Medicaid or other programs administered by CMS... Any use not authorized herein is prohibited, including... making copies of CPT for resale and/or license, transferring copies of CPT to any party not bound by this agreement, creating any modified or derivative work of CPT, or making any commercial use of CPT.
Read maximally literally, "internal use" and "CMS-administered programs only" would forbid nearly all third-party commercial reuse of ANY Medicare fee-schedule data — an interpretation the entire health-IT industry operates against daily (claims clearinghouses, EHR vendors, and billing software all reference CPT codes commercially). This pack's existing, already-shipped design applies the standard industry reading of the opening sentence — "CPT codes, descriptions and other data" are the copyrighted, licensed material; the CMS-computed derivative numbers indexed by those codes (RVUs, payment amounts, edit modifier indicators) are federal work product and are returned, while the AMA's own descriptive text is never stored or served. The PTP baseline's own "PTP Edit Rationale" column (a ~12-value CMS-authored categorical reason set, e.g. "Standards of medical/surgical practice") is not stored either, purely to keep this load a strict superset of the existing table shape (no schema change) — not because it reads as AMA content. This is the same theory already governing this pack's RVU/CLFS/MUE data; the PTP click-through does not introduce a new prohibition beyond what the rest of the pack already operates under, but the literal breadth of "internal use" / "CMS-administered programs" language is flagged here for anyone reviewing licence exposure across the pack.
Every PFS/CLFS source file opens with "CPT codes, descriptions and other data only are copyright American Medical Association". The RVU file scopes it precisely — codes and descriptions only — so the CMS-computed numbers are federal data and are returned, while the AMA's descriptive text is neither stored nor served. Callers pass a code they already hold.
DMEPOS is the exception, on purpose. It uses HCPCS Level II codes
(E/K/L/A-codes), which CMS itself maintains — unlike CPT (HCPCS Level I),
which is AMA copyright. The DMEPOS source file's own DMEREAD/DMEBACK
documentation carries no AMA/copyright notice anywhere (checked), so
medicare_dmepos_rate DOES return the code description.
Medicare pays DMEPOS by state (not by the physician-fee-schedule
locality), with a separate rural and non-rural amount per state. CMS's
own source file is one row per (HCPCS, modifier) with 53 state/territory
columns × 2 (non-rural/rural) = 106 rate columns; this pack keeps that row
grain and folds the 106 columns into one JSONB map rather than exploding to
~374,000 rows per release. medicare_dmepos_rate always returns both figures,
labelled, and says explicitly when a state carries no separately-adjusted
rural amount (a real "no adjustment" fact from the source file, not a missing
value) rather than presenting a bare null.
Loaded as of writing: 19,226 RVU rows, 109 localities, 2,081 lab rates, 3,531 DMEPOS code/modifier rows (2026 release C, 53 states/territories each), 33,354 MUEs, 4,109,749 distinct NCCI PTP pairs — the FULL CMS baseline (2026q3, v322r0) with the 2025q1..2026q3 quarterly change files layered on top (fleet #1293, 2026-09-07; superseded the prior 182,286-pair change-window-only coverage). All loaders refuse rather than reporting an empty parse as success.
ingest-dmepos.mjs resolves the current download from the DMEPOS release
subpage rather than guessing the zip filename — CMS's DMEPOS naming is not
stable (a release page 404s before it is posted, and old release URLs have
301-redirected to renamed paths), so it reads
.../dmepos-fee-schedule/dme<yy>[-<letter>] for the current .zip href on
every run, the same pattern ingest-medicare-pricing.mjs already uses for the
PFS RVU page.
ingest-ncci-ptp.mjs is additive/idempotent by design — re-running the full
default window is safe (upserts on service_type,column1,column2), and adding
a new quarter to --quarters widens the window forward without needing to
re-fetch history.
Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):
tools/list at https://gateway.pipeworx.io/medicare-pricing/mcp returns the tools in the table
above plus the shared Pipeworx meta-tools — ask_pipeworx,
discover_tools, search_within, remember/recall and the rest of the
gateway-wide set. So the tool count you see is larger than this table: a
single-pack endpoint currently lists roughly 30 shared tools alongside the
pack's own. The connection's initialize response states its exact scope, and
is the authoritative answer for a given day.
This is deliberate, not multiplexing by accident. The meta-tools are what let a
scoped connection answer a question this pack does not cover — via
ask_pipeworx, which routes across the whole catalog — without you adding a
second MCP server. There is currently no way to mount a pack endpoint without
them; if the extra schemas cost you more context than the routing is worth,
connect to the full gateway once rather than to several pack endpoints.
Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:
Both URLs reach the same gateway and the same 1683+ data sources. The
only difference is which pack's tools are listed directly; ask_pipeworx
reaches all of them from either one.
No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/medicare_physician_payment. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.
This package also runs as a local stdio MCP server — no Pipeworx account, no gateway round-trip:
Or run it directly to confirm it starts:
It speaks MCP over stdin/stdout and answers initialize/tools/list/tools/call
for only this pack's tools — none of the shared meta-tools the gateway
connection above adds. Same source, same tools, no ask_pipeworx routing.
Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:
The gateway picks the right tool and fills the arguments automatically.
MIT