Connect Stichii to your AI

Make your artwork wherever you like — ChatGPT, Claude, Midjourney, anything. Stichii turns it into a file your embroidery machine can actually sew.

Four ways in, easiest first. You only need one.

Paste a link

works with any chatbot

Nothing to install and nothing to sign up for. Any chatbot that can produce an image with a public link can hand you one of these, and it opens Stichii with your artwork already loading:

The link format

https://stichii.com/new?image=<your image URL>

Most chatbots won’t do this unless you ask, so paste this after your image request:

Ask any chatbot for a Stichii link

When you've made the image, upload it somewhere with a public https link and then give me a link in this exact form:

https://stichii.com/new?image=<the image URL>

I'll open that to turn the artwork into an embroidery file.

The image needs a public https link. If your chatbot only shows the picture in the chat, save it to your device and drag it into the editor instead — same result, one extra step.

Connect Claude Code

one command, needs an API key

Claude Code can talk to Stichii directly, so it can digitize artwork and hand you the finished file without leaving the conversation. Create an API key on the , then run:

Add Stichii to Claude Code

claude mcp add --transport http stichii https://stichii.com/api/v1/mcp \
  --header "Authorization: Bearer sk_stichii_..."

Not yet the claude.ai connector. Claude’s in-app Settings → Connectors dialog needs a sign-in-with-OAuth flow that Stichii doesn’t offer yet, so pasting this URL there won’t finish connecting. It’s on the list. Today this works in Claude Code and any client where you can supply an Authorization header — everywhere else, use .

You’ll still sign in to Stichii before any machine file is produced — previews and the readiness check are free, downloading a stitch file uses one of your exports.

Connect ChatGPT

more setup

ChatGPT reaches Stichii through the same server, using OpenAI’s connector/plugin support. Setup differs by plan and by what OpenAI has enabled for your account.

Server URL

https://stichii.com/api/v1/mcp

Being straight with you: OpenAI has changed how third-party tools plug into ChatGPT more than once recently (Actions → Apps → Plugins), and the current route expects a sign-in-with-OAuth flow Stichii doesn’t offer yet — so this one isn’t ready. Use — it works from ChatGPT on every tier and needs no setup at all.

Build on the API

for developers

A REST API and an MCP server, with API-key auth. Digitize images or text, get a fidelity score and a rendered preview, and export to 14 machine formats.

Endpoints

MCP      https://stichii.com/api/v1/mcp
OpenAPI  https://stichii.com/api/v1/agent/openapi.json
REST     https://stichii.com/api/v1/agent/designs

Full documentation, limits and auth details are on the .

Get artwork that sews well

Embroidery is thread, not pixels: a machine lays down solid blocks of colour with a needle. Anything that depends on a smooth blend, a soft edge, or a hairline detail has no thread equivalent, so it has to be approximated — and that approximation is what makes AI art sew badly.

Stichii cleans artwork up automatically, but the biggest single improvement happens before that — in what you ask for. Paste this into your image prompt:

Embroidery-friendly image prompt

Create the artwork as a flat vector-style design suitable for machine embroidery:

- flat vector illustration, like a screen-print or a printed logo
- bold, closed shapes with clean, continuous outlines
- a limited palette of 3-5 solid colours, high contrast between them
- absolutely no gradients, glow, bloom, drop shadows, or soft lighting
- no film grain, texture, noise, or photographic detail
- thick outlines; no hairline detail or tiny text
- plain solid background (or transparent), no scenery
- front-on, centred, symmetrical composition

Then give me a stichii.com/new?image= link so I can turn it into an embroidery file.

Why each rule

Flat vector style, like a printed logo or a screen-print — A needle cannot fade. Flat art already looks like what the machine can physically produce, so nothing has to be guessed at.
Bold, closed shapes with clean outlines — Each shape becomes a filled region the machine sews in one pass. Open or broken shapes leave the digitizer nothing definite to fill.
3–5 solid colours, no more — Every colour is a thread change — a stop, a re-thread, and a trim. Forty near-identical colours become forty stops and look muddy anyway.
No gradients, glow, bloom, or drop shadows — These have no thread equivalent. They get flattened into bands or noise, which is where "AI art turned to mud" usually comes from.
No film grain, texture, or noise — Grain reads as thousands of tiny colour regions, which becomes either a mess of stitches or a smear once simplified.
Thick outlines and generous minimum feature size — A satin column narrower than roughly 1 mm cannot hold together in thread. Thin lines drop out or pucker the fabric.
High contrast between adjacent colours — Thread colours that are close together stop being distinguishable once stitched, so the detail you paid a colour change for disappears.
Plain, solid background — or no background at all — A busy photographic background has to be either digitized (huge, ugly) or removed (better). Asking for a plain one saves the step.
Front-on, centred composition — Perspective and depth-of-field blur are pictorial tricks. Flat-on artwork maps directly onto a flat garment.

Platform details verified 2026-08-22. If something here has gone stale, and we’ll fix it.