Skip to main content
Get an API key, install the OpenAI client, and send a screenshot. The reply tells you what the model sees and what it would click. Model IDs, pricing, and tiers are on the introduction.
1

Get an API key

Generate a key on Portal-H and export it. The free tier needs no credit card. Add credits for higher limits and holo3-122b-a10b.
2

Install the OpenAI client

The API is OpenAI-compatible, so the official client works as is. Only the base_url changes.
3

Send a screenshot

Point the client at H, then send an image and a question.
Images can be a URL or a base64 data URI. Swap model for holo3-122b-a10b when you need maximum performance, or list what your key can reach with GET /v1/models.

Next steps

Agent loop

Multi-turn. Conversation and screenshots in, the next action out.

Element localization

Single-turn. A screenshot and a target description in, click coordinates out.

Document OCR

Single-turn. A page image in, its full text out.

API reference

Endpoints, request and response shapes, errors, and rate limits.
Last modified on September 11, 2026