Skip to main content

Pixy For Agents

Create automation key The Pixy MCP server lets an AI assistant use Pixy directly: find a template, read the elements it exposes, and render finished images or PDFs — without you writing the integration. It speaks Model Context Protocol, so any MCP client can connect: Claude, Cursor, VS Code, or your own agent.

What An Agent Can Do

Search the public template library by keyword, category or size, and read a template’s named elements to learn what can be changed.
Read your brand kits — logos, colors and fonts — and apply them to a render, so output matches your brand without anyone pasting hex values.
Apply text, image, fill and stroke changes to a design and get finished PNG, JPEG or PDF files back.
Render a design with its image element generated from a text prompt, when no photograph exists for what you are launching.

The Endpoint

Streamable HTTP, and stateless — there is no session to open and nothing is remembered between calls. Every request carries its own key.

Authentication

The same Automation key the REST API uses, as a bearer token. There is no separate OAuth flow to complete.
That is deliberate. A key you issue and can revoke is a smaller thing to hand an agent than a grant against your Pixy login, and it means the key in your MCP config is the same one your backend already uses — one credential, and one line per call in your request log.
Renders made by an agent are recorded with the source MCP, so you can tell them apart from your own backend’s calls when reviewing usage.

What It Costs

Exactly what the API costs, because the tools are the API: one credit per image page, one per PDF. Listing templates, reading a design and fetching brand kits are free. An agent is told to confirm with you before rendering in bulk, but it is worth knowing that generate_image spends credits every time it is called.

Where To Go Next

Get started

Connect Claude, Cursor or VS Code in four lines of configuration.

Tools

Every tool the server exposes, and what each one takes.