Luttie Now Has an MCP Server — Color Grade From Claude or Codex

Luttie now runs an MCP server. If you're working inside Claude Code, Claude Desktop, or Codex CLI, you can now ask your agent to browse the LUT library, generate an AI color grade from a description, or apply a LUT to an image — without opening a browser or leaving the conversation.

What Is MCP, Quickly

The Model Context Protocol is an open standard for connecting AI agents to external tools. Instead of an agent guessing at your app's API or scraping a webpage, it calls a defined tool with a defined input and gets a defined output back. Anthropic, OpenAI's Codex, and most modern agent clients support it.

Luttie's server is remote — hosted at luttie.app/api/mcp — so there's nothing to run locally and nothing to keep updated. You point your client at the URL with an API key, and the tools are just there.

Five Tools

• list_luts — every published LUT pack and single in the library, with slug, tags, and whether it's Pro. Free.
• get_lut — a LUT's metadata and .cube file contents by slug. Free LUTs return the file; Pro LUTs return a locked flag and upgrade link for free accounts.
• ai_grade — the same model behind Luttie's AI Grade feature, generating 3 grading variants from a text prompt and an optional reference image. Pro.
• apply_lut — applies a Luttie LUT to an image server-side, returning the graded result as a base64 PNG. No canvas, no editor session needed.
• account_status — your account's tier and current hourly usage for ai_grade and apply_lut, so an agent can check before it runs into a rate limit.

What This Actually Looks Like

Inside Claude Code or Claude Desktop, once the server's connected, this is a normal conversation turn:

"Grade this product shot warm and cinematic, then give me the LUT."

The agent calls ai_grade with your prompt and the image, gets three variants back, and can hand you the .cube file for the one you pick directly — or chain it into apply_lut to show you the graded result inline. No copy-pasting between a chat window and the editor.

It's also useful for LUT lookups without context-switching: "does Luttie have a Kodak Portra style LUT?" resolves to a list_luts call instead of you tabbing over to search the site.

Setup

Three steps, no local install:

1. Get an API key — sign in at Luttie, open your account menu, choose Developer / MCP, and generate one. It's shown once.
2. Add the server to your client. For Claude Code, the fastest path is the plugin:

/plugin marketplace add iwaju-labs/luttie-mcp-plugin
/plugin install luttie@luttie

Or connect directly from a terminal:

claude mcp add --transport http luttie https://luttie.app/api/mcp \
  --header "Authorization: Bearer YOUR_LUTTIE_API_KEY"

For Claude Desktop, it's Settings → Connectors → Add custom connector with the server URL and your key as a Bearer token. For Codex CLI, add the server to ~/.codex/config.toml and set LUTTIE_API_KEY in your shell.

3. Use it. Ask your agent to list LUTs, generate a grade, or apply one — it'll call the right tool.

Full config snippets for every client are on the MCP page.

Rate Limits and Tiers

ai_grade is capped at 20 generations per hour, apply_lut at 60 — tracked separately from the web editor's own limits, so using both doesn't cut into either budget. Free accounts can call list_luts, get_lut (free LUTs), and account_status with no restriction; ai_grade and Pro LUTs need a Pro, Lifetime, or Trial subscription.

Why We Built This

Luttie's editor is a browser tool, but a growing share of creative and technical work now happens inside an agent conversation — reviewing shots, drafting a look, scripting a batch export. Making the LUT library and AI grading callable as tools means that work doesn't need a context switch to a tab that isn't open.

Set up the MCP server →

Related

• Luttie MCP Server — full setup and tool reference
• AI Grade — generate LUTs from a text description
• Adjust your color grade by talking to it