Use QRdx with AI agents

Connect Claude, ChatGPT, Cursor, VS Code or Devin Desktop to the QRdx MCP server, give agents keyless QR image URLs, and call the REST API for a user.

AI assistants and coding agents can use QRdx in three ways. Pick by what the agent has to do:

The agent needs to	Use	Key or account
Make a QR code in a chat or an agent run	MCP server	None
Put a QR code in Markdown, HTML, an email or a build	Image API	None
Create or change dynamic codes, read scans	REST API	OAuth access token
Read the QRdx docs	Markdown and llms.txt	None

Connect the MCP server

The QRdx MCP server lives at:

https://mcp.qrdx.dev/mcp

It uses the Streamable HTTP transport and needs no account, sign-in or API key. The connect page has one-click buttons for Claude, Cursor and VS Code. The tools and their inputs are listed in MCP server.

Claude on the web and Claude Desktop share the same connectors.

Open Customize → Connectors and click Add custom connector. The Connect to Claude button on the connect page opens this dialog with the name and URL filled in.
Name it QRdx and enter https://mcp.qrdx.dev/mcp. Leave the OAuth fields empty.
Click Add, then turn QRdx on from the tools menu of a chat.

Free plans can add one custom connector. On Team and Enterprise plans, an owner adds the connector for the organization.

Then ask in plain words, for example "Make a QR code for https://example.com in #155dfc with rounded dots" or "Make a Wi-Fi QR code for the network Office Guest".

Discovery for MCP clients

Clients and registries that discover servers on their own can read:

the MCP Server Card at https://mcp.qrdx.dev/mcp/server-card, in the format of the MCP Server Card extension: name, version, transport and supported protocol versions;
the AI Catalog at /.well-known/ai-catalog.json, which points to the server card;
the API catalog at /.well-known/api-catalog (RFC 9727), which lists the REST API with its OpenAPI document and the MCP server.

Give agents QR image URLs

The image API turns a URL into a QR code image with no key, so an agent can write the image straight into Markdown, HTML or an email:

https://api.qrdx.dev/v1/qr?data=<encoded text>&format=svg|png|eps

![QR code for qrdx.dev](https://api.qrdx.dev/v1/qr?data=https%3A%2F%2Fqrdx.dev&format=png&size=512)

Encode data with encodeURIComponent, or &, #, + and spaces split it.
Pick the format for the destination. PNG works in chat apps and email clients that don't show SVG; SVG is sharpest on the web; EPS is for print shops.
Style with query parameters (fg, bg, body, eye, dot, level, margin, template). The values are listed in Image API, and the MCP tool list_qr_styles returns them too.
The same URL always returns the same image, so it's safe to cache and embed.

Call the REST API for a user

To create or change dynamic codes, read scans or subscribe to webhooks, an agent calls the REST API at https://api.qrdx.dev/v1 with an OAuth 2.0 access token. The OpenAPI document describes every endpoint, schema and error, and the API reference has a page for each.

QRdx doesn't issue API keys yet, so an agent gets a token the way any OAuth app does:

Register an app in Settings → Developers with the scopes it needs and a redirect URI it can receive, such as a http://localhost callback for a command-line agent.
Run the authorization code flow with PKCE once. The user approves the app on the QRdx consent screen; the agent exchanges the code for an access token and a refresh token. The steps are in Get an access token.
Keep the refresh token in a secret store, never in a prompt, a log or the repository. Access tokens last an hour; refreshing returns a new refresh token and the old one stops working.
Ask for the least access. codes:read is enough to report on codes and scans; codes:write creates and edits them.
Respect the quota. Each workspace has a daily request quota by plan (see Plans and limits). On 429 RATE_LIMITED, wait for the seconds in Retry-After.

Requests made with a token act on the token owner's personal workspace, not on a team.

Read QRdx as Markdown

/llms.txt: an index of the whole site for LLMs, with the product, developer tools, docs, help center, pricing and legal pages.
/llms-full.txt: every docs page in one file, in sidebar order.
Add .md to any docs URL for that page alone, for example /docs/developers/mcp.md. /pricing.md and /mcp.md work too.
Requests to the home page, /pricing, /mcp or any docs page with Accept: text/markdown get the Markdown version at the same URL.

Limits

The MCP tools make static QR codes. They render images and build payloads (Wi-Fi, contacts, email, SMS, map locations). They can't create dynamic codes or read anything from a QRdx account; use the REST API for that.
No logos or gradients from the MCP server or the image API: they draw solid colors. Use the playground or the React library for logos and gradients.
Image URLs contain the encoded data. Don't put passwords or other secrets in them, such as a private Wi-Fi password in a public README.
Rate limits apply per IP address to the MCP server and the image API; the numbers are on the MCP server and Image API pages.
REST API tokens come from OAuth only. There are no API keys and no client-credentials grant yet, and tokens can't act on team workspaces.

Use QRdx with AI agents

Connect the MCP serverCopy Anchor Link

Discovery for MCP clientsCopy Anchor Link

Give agents QR image URLsCopy Anchor Link

Call the REST API for a userCopy Anchor Link

Read QRdx as MarkdownCopy Anchor Link

LimitsCopy Anchor Link

On this page