MCP App Store
by KeenEthics
PenguinAI Medical Coding icon

PenguinAI Medical Coding MCP Server for ChatGPT

By Penguin AI

Overview

PenguinAI Medical Coding turns de-identified medical charts into coding suggestions a professional coder can review. - ICD-10-CM diagnosis codes with CMS-HCC v24, v28 and RxHCC mappings, each backed by MEAT evidence and a verbatim citation from the chart - deterministically from the medical decision making elements, with the 2023 AMA rationale - Clinical summaries: diagnoses, medications, labs, vitals and timeline, with citations - Accept or reject suggested codes, attributed to the signed-in coder De-identified documents only: every upload is screened for patient identifiers, and charts that contain them are refused. Results are returned de-identified (encounters labelled by year). Sign in with your Gwen account (gwen.penguinai.co). Usage is charged to your Gwen credits at actual processing cost. Suggestions are not billing determinations and must be reviewed by a qualified coder.

How to connect

  1. ChatGPT: Open the PenguinAI Medical Coding listing in the ChatGPT app directory and choose Connect. ChatGPT listing PenguinAI Medical Coding exposes 9 tools here.

Tools

calculate_em_level

ChatGPT
Calculate the E/M CPT code from MDM element levels the user has already decided, or from total encounter time. This is a fixed table calculation; no note is needed. Provide either the three MDM elements (problems, data, risk, each MINIMAL, LOW, MODERATE or HIGH; the level is the highest met by at least two) or total_minutes for time-based selection. Time-based selection does not apply to the emergency department. The result is a suggestion for review by a qualified coder, not a billing determination.

code_em_note

ChatGPT
Suggest an E/M (Evaluation and Management) visit level for a de-identified clinical note, using the 2023 AMA medical decision making (MDM) rules. Each MDM element (problems, data reviewed, risk of management) is rated MINIMAL, LOW, MODERATE or HIGH with a written rationale, and the level is the highest one met by at least two of the three elements. setting is the place of service, one of: ed, office_new, office_est, inpatient_initial, inpatient_subsequent, nf_initial, nf_subsequent, home_new, home_est, consultation_office, consultation_inpatient, telehealth_new, telehealth_est. get_em_reference shows the tables. Only de-identified notes are accepted; a note that contains patient identifiers is refused. The result is a suggestion for review by a qualified coder, not a billing determination.

code_home_health_episode

ChatGPT
Suggest home health episode coding for a de-identified chart or note: a principal diagnosis with its PDGM clinical group, sequenced secondary diagnoses, and procedure codes for documented procedures, applying home health coding rules (for example suppressing codes that cannot be primary, 7th-character D for injuries, and visit-type sequencing). Provide the chart as chart_file (the file the user attached), pdf_base64, or note_text. Optional focus_of_care and visit_type (aftercare, injury, disease_management, wound, sequela, device) come from the referral; otherwise they are inferred from the chart. Only de-identified charts are accepted; a chart that contains patient identifiers is refused. Results are suggestions for review by a qualified coder, not billing determinations.

get_chart_results

ChatGPT
Fetch the coding results for a chart that was already uploaded: ICD-10 codes grouped by Date of Service with HCC mappings (CMS-HCC v24/v28, RxHCC) and MEAT evidence. Use the document_id returned by process_chart. Does not re-upload the chart. If the chart has not finished coding yet, this reports status="processing" rather than returning an empty result — an unfinished chart is never mistaken for one with no diagnoses.

get_chart_status

ChatGPT
Wait for a chart to finish coding. Returns state=done|processing|error. This BLOCKS for a short while waiting for progress, so you do not need to (and cannot) sleep between polls — just call it again if it still reports state="processing". Once it reports state="done", call get_chart_results to get the codes.

get_em_reference

ChatGPT
The E/M reference tables: which settings exist, how each MDM level maps to a CPT code, and the time thresholds. Use this to pick a valid setting, or to show a coder the rule behind a level rather than asking them to trust it.

process_chart

ChatGPT
Suggest ICD-10-CM diagnosis codes for a de-identified medical chart PDF, with MEAT evidence and a verbatim citation for each code, mapped to CMS-HCC v24, CMS-HCC v28 and RxHCC categories. Codes are grouped by encounter. Provide the chart as chart_file (the file the user attached) or pdf_base64. member_id is a pseudonymous label chosen by the user to group charts for the same patient (for example DEMO-1); it must not be a real member, beneficiary or medical record number. Only de-identified charts are accepted. A chart that contains patient identifiers is refused and is not coded. Coding can take a few minutes: if the result is not ready, this returns status="processing" and a document_id for get_chart_status and get_chart_results. Results are coding suggestions for review by a qualified coder, not billing determinations.

submit_code_decision

ChatGPT
Record a coder's review of one AI-suggested code: decision must be "accepted" or "rejected". Use the code_id from the chart's results. Reversible — submitting the opposite decision later overrides it.

summarize_chart

ChatGPT
Summarize a de-identified patient chart: active and historical diagnoses (with status), medications (with regimen and status), lab results, allergies, vitals, an encounter timeline and a short narrative. Each item includes the verbatim line from the chart it came from. Provide the chart as chart_file (the file the user attached) or pdf_base64, or pass already-transcribed text as note_text. This tool does not assign billing codes. Only de-identified charts are accepted; a chart that contains patient identifiers is refused.

Capabilities

Writes

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/penguinai-medical-coding"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/penguinai-medical-coding)