# Healthie Dev Assist
**MCP server for AI-powered Healthie API exploration**
[](LICENSE)
[](https://nodejs.org)
[](https://modelcontextprotocol.io)
*From 5–10 tool calls to 1–2. Schema-aware. Sandboxed. Instant.*
---
---
## Overview
Classic MCP servers expose one operation per tool call — search a type, wait, introspect it, wait, build a query, wait. **Dev Assist 2.0** changes this with a **code execution model**: your AI writes a small TypeScript program that performs all steps in a single turn.
| | v1 | v2 |
|---|:---:|:---:|
| Turns per typical task | 5–10 | **1–2** |
| Schema exploration | One call at a time | Search + introspect + query in one shot |
| Tools exposed | Many | **One** (`execute_healthie_code`) |
| Schema lookups | Live API each turn | Cached locally — instant |
**Example**: "Find all appointment mutations and show me what `createAppointment` takes" — previously 3+ tool calls. In 2.0:
```typescript
const mutations = await healthie.search("appointment", { kind: "mutation" });
const details = await healthie.introspect("createAppointmentInput");
return { mutations, details };
```
Done in one execution.
---
## API
Your AI has access to a single `healthie` object inside a sandboxed Node.js environment (no filesystem, network, or shell access — only these methods):
```typescript
// Search the schema by keyword
healthie.search(query: string, options?: { kind: "query" | "mutation" | "type" }) → SearchResult[]
// Get full details on any type, query, or mutation
healthie.introspect(typeName: string, options?: { depth: number }) → TypeDetails
// Execute a GraphQL query
healthie.query(graphql: string, variables?: Record