# AuthWeaver MCP

AuthWeaver is an MCP server tool that extracts prior-authorization evidence from synthetic FHIR data and returns a structured evidence packet for clinicians.

## What It Does
- Receives SHARP context (patient ID + optional FHIR server URL + token)
- Queries a synthetic FHIR server for relevant resources
- Uses an LLM to extract medical-necessity evidence
- Returns a strict JSON evidence packet

## Safety & Privacy
- Stateless: no patient data stored
- Data-minimized FHIR queries
- Output is structured for audit review
- Synthetic/de-identified data only

## Quick Start (Local)
1. `python -m venv venv`
2. `venv\Scripts\activate` (Windows) or `source venv/bin/activate` (macOS/Linux)
3. `pip install -r requirements.txt`
4. Copy `.env.example` to `.env` and set an LLM key
5. `python server.py`

## MCP Tool
Tool name: `extract_prior_auth_evidence`

Inputs:
- `procedure_name` (string)
- `sharp_context` (object, optional)
- `patient_id` (string, optional, dev only)

Output (JSON):
- `procedure`
- `criteria_met` (list of evidence items)
- `criteria_not_met` (list)
- `clinical_summary`

## Demo Flow
1. Prompt Opinion agent invokes `extract_prior_auth_evidence` with SHARP context.
2. AuthWeaver pulls synthetic FHIR data and extracts evidence.
3. Agent displays the evidence packet in the clinician workflow.

## Notes
- The MCP SDK wiring may differ based on Prompt Opinion docs. This repo includes a best-effort MCP server plus a FastAPI fallback for local testing.
- Use only synthetic data (Synthea or public FHIR sandbox).