Ask Claude, Cursor or VS Code for colors with the Tintly MCP server
Adding Tintly to an assistant takes about two minutes and two ingredients: a free API key and the server address. After that, the assistant can generate palettes, measure contrast, explain what a color means and preview a palette for color blindness, all from the chat or the editor you already use.
- 1Create an API key
Open the developer page, sign in and press Create key. Name it after where you will use it, such as Claude Code on my laptop. The key is shown once, so copy it straight away.
- 2Open your assistant's MCP settings
Every assistant keeps its servers in a settings screen or a JSON file. The setup guides below name the exact place for each.
- 3Paste the address and the key
The address is the same for everyone. The key goes in an Authorization header as a Bearer token.
- 4Ask a first question
Try: make me a triadic palette from #2463EB and check which pairs are readable as text. The assistant should call Tintly's tools and show real contrast ratios.
The shortest setup: Claude Code
Claude Code can add a remote server with one command. Replace YOUR_API_KEY with your key and run it in a terminal.
claude mcp add --transport http tintly https://www.trycolorkit.com/api/mcp \
--header "Authorization: Bearer YOUR_API_KEY"Run /mcp inside Claude Code to confirm that Tintly is connected and to see its tools.
Editors that read a JSON file: Cursor and VS Code
Cursor reads mcp.json and VS Code reads .vscode/mcp.json. Cursor takes the key in a headers block. VS Code can ask for the key when the server starts, which keeps the secret out of a file you might commit.
{
"mcpServers": {
"tintly": {
"url": "https://www.trycolorkit.com/api/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}{
"inputs": [
{ "type": "promptString", "id": "tintly-key", "description": "Tintly API key", "password": true }
],
"servers": {
"tintly": {
"type": "http",
"url": "https://www.trycolorkit.com/api/mcp",
"headers": { "Authorization": "Bearer ${input:tintly-key}" }
}
}
}More assistants
Claude Desktop
Uses the small mcp-remote bridge from claude_desktop_config.json, because desktop config starts local programs. Needs Node.js.
Windsurf and Cline
Both read a JSON settings file. Windsurf uses serverUrl, Cline uses type streamableHttp, and both accept a headers block.
Codex CLI and Gemini CLI
Codex reads a TOML file and can take the key from an environment variable. Gemini CLI uses httpUrl in settings.json.
Which assistant should you start with?
Start with the one you already have open all day. If that is an editor, the editor guides are the quickest because the configuration sits next to your code. If you mostly chat, Claude Desktop works well, though it needs Node.js for the small bridge it uses. A command-line agent is the quickest of all to try, because the setup is one command. You can add Tintly to more than one assistant later, with a separate key for each so that each can be revoked on its own.
Keep the key safe
- Do not commit a file that contains a real key. Use an environment variable or an input prompt where your assistant supports one.
- Make one key per place you use it, so you can revoke a single laptop or project without breaking the others.
- If a key leaks, revoke it on the developer page. It stops working within a minute.
When something does not connect
| You see | It usually means | Fix |
|---|---|---|
| 401 or invalid key | The key is mistyped, revoked or expired | Create a new key and paste it again, with no spaces or quotes around it |
| 429 or rate limited | More than 60 requests a minute on one key | Wait for the Retry-After seconds, or use a separate key for the busy tool |
| Server listed but no tools | The assistant did not finish the handshake | Restart the assistant and check its MCP log for the error |
| Method not allowed on GET | The client tried to open a stream | This is expected: Tintly answers POST requests only |
Every Tintly tool is read-only, so you can let your assistant call them without approving each one. When you are ready for more, the next article covers the same tools through a plain HTTP API for your own code.
Sources and further reading
- Transports (Streamable HTTP) · Model Context Protocol