---
name: humiris-integration
description: Use when adding Humiris managed AI models, routing, MCP, gateway or observability to an application.
---

# Humiris integration

Integrate Humiris as a small, testable server-side adapter.

## Required behavior

1. Inspect the repository, runtime, package manager and existing AI client before editing.
2. Reuse the official OpenAI SDK if it is already present. Otherwise add it using the project's package manager.
3. Read the API key from `HUMIRIS_API_KEY`. Never hardcode, print, log or send the key to browser code.
4. Use `https://api.humiris.ai/api/v1` as the OpenAI-compatible base URL.
5. Default to `humiris/auto`, while keeping the route configurable.
6. Put provider access behind one adapter so the rest of the application is not coupled to an SDK.
7. Preserve existing error handling, types, style and user experience.
8. Add and run a smoke or integration test that exercises the adapter.

## MCP

The public Humiris developer MCP is available at `https://humiris.ai/api/mcp`. Use it to search models, recommend routes and generate starter integrations.

## Completion

Report changed files, checks run, their exact results and the environment variable the developer must configure. Do not claim success unless the relevant build and test pass.

References: https://humiris.ai/developers and https://humiris.ai/get-started