MPHI: a statewide rate platform built with Claude Code, queried in Claude
The Michigan Public Health Institute sat on a large corpus of claims and public health data it could not put to work. With Aequalis, a payer rate platform one developer built with Claude Code, 3Pillars loaded every published BCBS Michigan and HAP rate, ran a behavioral health and substance use rate assessment, and gave MPHI its own seat on an MCP server so its people query the data in Claude.
- Claude in production
- Claude Code (the platform and its MCP server); Claude API through the Aequalis AI Assistant; Claude Desktop over MCP; Claude Sonnet and Opus
- Status
- Production deployment Q4 2025 to Q1 2026; MPHI on the MCP server today
- Data
- Every published negotiated rate for BCBS Michigan and HAP, about 10 million rates
- One developer built the Aequalis platform with Claude Code in about a month, against about six months for a conventional team build; about 2 TB of published payer files became 11.9 million enriched rates.
- Every published BCBS Michigan and HAP rate was reviewed; behavioral health and substance use services were pulled out and compared. The main finding: MPHI could create a novel tool to identify cost savings.
- The deployment replaced up to 4 FTEs at MPHI, $600K to $800K a year, and MPHI now queries the live data in Claude on its own MCP seat with no data engineering, SQL or exports.
The problem we took
MPHI is a nonprofit that works with government, academia, community groups and providers on public health programs. It holds proprietary Medicaid and Medicare claims and Michigan public health data, and it could not put that corpus to work. What it lacked was utilization and pricing: what Michiganders actually use across the healthcare system around them, and what that care costs. Federal rules make health plans publish their negotiated rates, but the files are enormous, inconsistent and not built for analysis: single payer files run from 5 to 200 GB, and a statewide read means more than 100 million raw rate rows.
MPHI wanted one structured, trustworthy view that is useful to the public and that it can commercialize for plans, providers, specialty providers and health systems, without building or staffing a data platform of its own.
What the client already had
The corpus, the public health mission, and a business plan with seven customer types from public good to paid: population health and reform, patients, researchers, insurance departments, private payers and providers, employers, and commercial vendors, sold as free public dashboards, paid canned reports and paid specialized datasets. MPHI paid for the initial specification; 3Pillars delivered beyond it.
What we built
Aequalis, 3Pillars’ payer rate data platform: the ingest pipeline, the enrichment and confidence scoring, the search index and the API behind it, built by one developer with Claude Code. From Q4 2025 through Q1 2026, 3Pillars loaded every published Blue Cross Blue Shield of Michigan and HAP rate and ran a production deployment for MPHI, working the data through the Aequalis AI Assistant (Claude via the Anthropic API) and in Claude with rate extracts.
Then an MCP server for Aequalis, again built with Claude Code, that exposes the platform’s governed endpoints as tools inside Claude Desktop and Claude Code, with per-payer routing, scoped API keys and role-based access. It went live in June 2026. MPHI has its own seat and uses it today: its people ask their questions in plain language and get answers from the live rate data.
What it returned
A read of every published rate for both plans, with behavioral health and substance use compared across plans and provider groups: on the professional-rate medians for seven codes, one plan ran higher on five of the seven by 10% to 71%, and substance use codes were sparse in both files. MPHI paid for the initial specification; 3Pillars delivered beyond it, and what runs today is a live production environment in daily use. Aequalis holds every published BCBS Michigan and HAP rate, MPHI’s people query it in Claude on their own MCP seat, and the outputs 3Pillars laid out in the February 2026 overview run from that live data across the seven customer types in MPHI’s own business plan, from public good to paid: a quarterly affordability benchmark book, regional access dashboards for behavioral health and other key specialties, and a watchlist of high-variation services with the savings opportunity sized.
Claude changed what MPHI could do with its own data, and MPHI needed no software of its own.
Savings on the public health side are not measured yet. The live platform is at aequalishealth.ai.
Run this inside your organization.
A 30-minute working session. We walk through what an engagement like this one would look like for you.