# VerifiedX LangGraph single-agent healthcare starter

Protect a healthcare prior-auth graph before a single agent writes the wrong downstream state.

This starter mirrors a real healthcare RCM flow:

- the graph looks up a prior-auth case
- a safe workqueue action is allowed
- an unsafe attempt to mark the authorization approved is blocked
- the application catches the block and routes to manual review so work keeps moving

It is intentionally model-free so the single-agent LangGraph boundary stays obvious before you plug in your own model loop.

## Run it

```bash
python -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt
verifiedx doctor
```

Set env:

```bash
export VERIFIEDX_API_KEY=...
export VERIFIEDX_BASE_URL=https://api.verifiedx.me
```

Run:

```bash
python run.py
```

What to check:

- `artifacts/scenario_report.json`
- `artifacts/verifiedx_diagnostics.jsonl`

You should see an allowed workqueue write, a blocked attempt to mark the authorization approved in the EMR, and a manual review fallback that keeps the case moving.