SVGMill for Agents and MCP

A hosted MCP server that gives your agent one job it can do well: turn an image into a vector file. It uses your API key and your credit balance.

The server

The MCP endpoint is https://svgmill.com/mcp. It speaks Streamable HTTP with plain JSON responses, so any client that supports remote HTTP servers can connect. Authenticate with your API key as a Bearer token, the same key the REST API uses.

Tool descriptions are written in plain English and say what each call costs, so the agent can explain a charge before it makes one.

Three tools

Enough for an agent to convert, fetch other formats, and check that it has credit left.

ToolTakesReturnsCost
convert_image_to_svgimage_url or image_base64, and optional format, colors, presetThe file (SVG as text, other formats as a base64 resource), its token and the credits left1 credit
download_vectortoken, formatAnother format of an earlier traceFree for a paid trace; 1 credit the first time a preview token is downloaded
get_accountNothingBalance, plan and usage this periodFree

Set it up in your client

Replace svgmill_live_… with your key. Keep keys out of shared repositories.

claude mcp add --transport http svgmill https://svgmill.com/mcp \
  --header "Authorization: Bearer svgmill_live_…"

No MCP? Call the API

Agents that can make HTTP requests can use the REST API directly. One call converts the image and returns the file:

curl https://svgmill.com/v1/vectorize \
  -H "Authorization: Bearer svgmill_live_…" \
  -H "Content-Type: application/json" \
  -d '{"image_url":"https://example.com/logo.png","format":"svg"}' \
  -o logo.svg

The full reference is in the API docs. For reading by machines there is llms.txt, llms-full.txt and the OpenAPI 3.1 spec.