Give your AI agent color grading tools.

Luttie's MCP server connects Claude, ChatGPT, Codex, and any MCP-compatible agent straight to Luttie. List LUTs, generate a grade from a text prompt, and apply a LUT to an image, all without leaving your agent.

What is the Luttie MCP server?

MCP (Model Context Protocol) is an open standard that lets AI agents call external tools directly. Luttie runs a remote MCP server at luttie.app/api/mcp that exposes the LUT library, AI Grade, and LUT application as callable tools. Point Claude or Codex at it with your API key, or connect ChatGPT over OAuth, and your agent can browse LUTs, generate a look from a description, or grade an image mid conversation.

Setup

It's a remote server, so there's nothing to install. Add the URL to your client's MCP config, with an API key for Claude/Codex or an OAuth sign-in for ChatGPT.

1. Get your API key

Sign in at Luttie, open your account menu, and choose Developer / MCP. Generate a key there. It's shown once, so copy it before closing the dialog.

2. Claude Code

Easiest: install the Luttie plugin. It sets up the connection for you, you just need to export your API key first:

terminal
export LUTTIE_API_KEY="YOUR_LUTTIE_API_KEY"
claude code
/plugin marketplace add iwaju-labs/luttie-mcp-plugin
/plugin install luttie@luttie

Or skip the plugin and connect directly, either from a terminal:

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

Or by adding it to your project's .mcp.json directly:

.mcp.json
{
  "mcpServers": {
    "luttie": {
      "type": "http",
      "url": "https://luttie.app/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_LUTTIE_API_KEY"
      }
    }
  }
}

3. Claude Desktop

Open Settings > Connectors > Add custom connector, paste in the server URL below, and add your API key as a Bearer token when prompted:

server url
https://luttie.app/api/mcp

4. Codex CLI

Codex doesn't yet have a CLI command for remote servers, so add this to ~/.codex/config.toml by hand, then set LUTTIE_API_KEY in your shell environment:

config.toml
[mcp_servers.luttie]
url = "https://luttie.app/api/mcp"
bearer_token_env_var = "LUTTIE_API_KEY"

5. ChatGPT

ChatGPT connects over OAuth, not an API key, no copying keys around. In ChatGPT, turn on Settings > Apps & Connectors > Advanced settings > Developer mode, then create a new connector with this server URL and Authentication: OAuth. You'll be redirected to sign in to Luttie and approve access.

server url
https://luttie.app/api/mcp

Available tools

Five tools, covering the LUT library, AI grading, and image processing.

list_lutsFree

Lists every published LUT pack and single in the Luttie library, with slug, title, tags, and whether it needs Pro.

get_lutFree / Pro

Fetches a LUT's metadata and .cube file contents by slug. Pro LUTs return a locked flag and upgrade link for free accounts.

ai_gradePro

Generates 3 color grading variants from a text description and an optional reference image, using the same model behind Luttie's AI Grade feature. When you supply an image, each variant comes back with a preview of it graded.

export_lutPro

Renders a grade to a real .cube or .3dl file and returns a 24-hour download link. Pass an ai_grade variant's settings straight in — this is what turns an AI grade into a LUT you can load in Resolve, Premiere, or CapCut.

apply_lutFree / Pro

Applies a Luttie LUT to an image server side and returns the graded result as a base64 PNG, no browser required.

account_statusFree

Returns the authenticated account's tier and current hourly usage for ai_grade, export_lut, and apply_lut.

Frequently asked

It's a remote server that speaks the Model Context Protocol (MCP), an open standard for connecting AI agents to external tools. It exposes Luttie's LUT library, AI color grading, and LUT application as tools an agent can call directly, without a browser.

No. It's a remote MCP server, so you just add its URL and your API key to your client's MCP config. There's nothing to run locally and nothing to keep updated.

Any MCP client that supports remote servers over Streamable HTTP, including Claude Code, Claude Desktop, ChatGPT (Developer Mode connectors), and Codex CLI. If your client only supports local (stdio) MCP servers, check its docs for a remote-server bridge.

Listing LUTs, reading free LUT files, and applying free LUTs to images work on any account, including free ones. AI Grade and Pro-only LUTs need a Pro, Lifetime, or Trial subscription, the same as the web editor.

Sign in at luttie.app, open your account menu, and choose "Developer / MCP." Generate a key there. It's shown once, so copy it into your MCP config right away. ChatGPT doesn't need this step, it connects over OAuth instead.

ChatGPT only supports OAuth for MCP connectors, not static API keys, so it signs in through Luttie directly instead of using a generated key. Claude and Codex use the API-key path since that's what those clients support. Both paths hit the same server and the same tools.

Yes. AI Grade is capped at 20 generations per hour and LUT application at 60 per hour, tracked separately from the web editor's own limits so the two don't interfere with each other.

Connect your agent

Sign in, generate a key, and add Luttie to your MCP config. Free accounts can browse and apply free LUTs right away.