About MedCoder.ai

MedCoder.ai is a free, version-aware reference and decision-support system for the published US medical code sets and the public CMS rules built on them: ICD-10-CM diagnoses, HCPCS Level II procedures and supplies, ICD-10-PCS inpatient procedures, MS-DRG payment groups, clinical laboratory reference pages, and the CMS-derived coding rules — procedure-to-procedure edits, published utilization limits, National Coverage Determination code lists and MS-DRG grouping — that apply to a whole claim. Every rule it reports names the source it came from and the release that source was published in.

Current data releases:ICD-10-CM FY2027 · ICD-10-PCS FY2027 · HCPCS October 2026 · MS-DRG v44 · Medicare Code Editor v44.0 · NCCI PTP Q4 2026 · MUE Q4 2026 · NCD code lists 2026-01 · LCD export September 28, 2026 · All releases and sources

What MedCoder.ai is for

MedCoder.ai is built on a single principle: medical coding information should be connected, not isolated.

The conventional coding reference has barely changed in years — search for a code, open a page, read the description, move on to the next code. But a code is never an isolated fact. It carries Excludes1 and Excludes2 relationships, inclusion terms, additional-code instructions, historical changes, risk-adjustment implications, and connections to entirely different coding systems. A reference that shows only the description leaves the coder to reconstruct all of that by hand.

A connected coding reference

Rather than a collection of independent pages, MedCoder.ai models the relationships between the information itself — ICD-10-CM connected to ICD-10-PCS, HCPCS, MS-DRG, clinical classifications, and laboratory information. That structure is what makes the platform’s investigative and decision-support tools possible: code comparison, codes by condition, coding history, Claim Check, and natural-language search.

The visible website is the smaller half of the project. The larger half is the information architecture underneath it: structured data, reliable relationships, version awareness across fiscal years, historical change tracking, and an architecture that handles hundreds of thousands of pages and millions of relationships.

How the data is verified

Each registry is loaded from a named official release file. The Data Sources page records that file, the publisher’s download link and the SHA-256 checksum of the copy MedCoder loaded, so anyone can fetch the same file and check it. Audit scripts in the site’s build re-parse those files — the ICD-10-CM order files, the ICD-10-PCS order file, the HCPCS release, the MS-DRG definitions and their severity, diagnosis and procedure appendices, the CMS-HCC mappings, the Medicare Code Editor age and principal-diagnosis edits, the AHRQ clinical classifications and the Medicare lab coverage lists — and compare every stored code, title, billable flag, instructional note and relationship against them, at each ingest and before a release goes live. The CMS rules data behind Claim Check is loaded the same way, release by release: the NCCI procedure-to-procedure and Medically Unlikely Edit tables, the National Coverage Determination manual and its published diagnosis code lists, the HCPCS Level II modifier registry and the MS-DRG grouper package, each with its own audit script and its own recorded checksum.

That is MedCoder’s own process, not an independent certification. The official files remain the authority; every page links back to them and quotes the official text where it applies, so a reader need not take the comparison on trust.

The MS-DRG grouper is checked against CMS’s own software on a stated fixture corpus, release by release, with the exact-match results, the mutation testing, the unresolved differences and the limitations recorded: How MedCoder verifies its implementation.

Editorial policy on generated content

No coding fact on MedCoder.ai is AI-generated. Every code, title, billable status and instructional note comes from the official release files described under “Where the data comes from” below, and a page exists only where a registry row exists. Where a fact is not published by a federal source, the site says so instead of supplying an estimate — the laboratory reference ranges below are the clearest example.

Some relationships are computed, not transcribed: MedCoder derives them deterministically from those same datasets — laterality pairs matched from code titles, or an MS-DRG grouping run against the CMS grouper tables — and labels them MedCoder-derived or MedCoder-calculated wherever they appear. A grouping is a MedCoder implementation of CMS’s published logic, not a run of CMS’s own grouper software: MedCoder is not certified by, endorsed by, approved by or affiliated with CMS. Figures that could not be verified against a source are omitted rather than approximated, and a release that is not loaded is reported as data unavailable rather than as a clean result.

The site's editorial layer — per-code FAQ answers, the coding guides, the curated comparison explanations (the comparison verdicts themselves are MedCoder-derived from the tabular notes) and release narratives — is drafted with AI assistance under a written sourcing prompt that requires every factual claim to trace to a specific official release file, then verified by a person against that file before publishing. That is a distinct process from the coding facts above: an editorial sentence is reviewed prose, not a transcribed CMS/CDC value, and is never treated as a substitute for the tabular notes it explains.

Where the data comes from

Codes, titles, billable status and the official tabular instructional notes (Includes, Excludes1, Excludes2, Code First, Code Also, Use Additional Code) are taken from the release files for the current fiscal year — ICD-10-CM from CMS and the CDC; HCPCS Level II, ICD-10-PCS and MS-DRG from CMS. These are works of the US federal government. The exact release version behind every data set is listed on the Data Sources page.

The laboratory pages are the one section not built from the code files, so they carry their own sources. Test descriptions come from MedlinePlus, National Library of Medicine — also a public-domain US government work, credited on every page that uses it. Reference ranges are published only where a federal health agency states the value, and each range links to the page it came from (NIDDK, the NIH Office of Dietary Supplements, and the National Cancer Institute). Most laboratory pages therefore show no reference range at all: an interval is a property of the performing laboratory and its assay, so there is no national figure to cite, and inventing one would break the rule above.

Scope and limitations

MedCoder covers ICD-10-CM, ICD-10-PCS, HCPCS Level II and MS-DRG, and the CMS rules built on them that it holds. It does not include CPT codes, descriptors or modifiers, or other AMA-licensed CPT content, so its NCCI procedure-to-procedure and Medically Unlikely Edit information is the supported HCPCS Level II subset, not the complete tables. Local Coverage Determinations are identified for review, not evaluated; National Coverage Determination checks compare codes with the code lists CMS publishes; and coverage, payment and medical-necessity determinations remain with the payer. Payer-specific edits, Outpatient Code Editor logic and fee schedules are not modeled.

The Scope and Limitations page sets out each boundary, and what results such as “data unavailable” and “no issue found” mean.

See the Privacy Policy and Terms of Use.