imagic Max ships with an MCP server and a scriptable CLI, so your AI agent can drive the same culling, editing, and export pipeline you'd run by hand. Point Claude Code, Claude Desktop, Cursor, Codex, or Copilot at your photo library and give it a prompt.
Included in imagic Max, €99 once. Not sure yet? The free trial unlocks it with no card required.
Every tool below is the same automation imagic's headless CLI has offered for a while, now reachable by any MCP client.
imagic's MCP server works with any MCP client. Requires an activated imagic Max license or an active trial. Here's how to connect the ones people ask about most.
If you installed imagic via pip (MCP is an optional extra):
If you installed the packaged Windows app:
Add this to your claude_desktop_config.json:
Most MCP clients that read a JSON config file, including Cursor, accept the same shape. Add it to that client's MCP config:
No AI client needed. imagic's headless mode has been scriptable since before MCP existed:
Included in imagic Max, €99 once. MCP and the CLI aren't a separate add-on, they're part of the top tier alongside style calibration, neural denoise/upscale, and semantic search.
Free 7-day trial includes MCP and the CLI. No card required, try the full agent workflow before you decide on a tier.
Automation is only useful if the output looks like you edited it.
The first time you connect an AI agent, imagic runs a short guided calibration. It pulls up to 8 scenario photos from your own library (fewer if your library does not cover every scenario type yet), covering a spread of lighting and subject types, and asks you to edit them the way you normally would.
It watches the exposure, color, and crop choices you make on those photos and uses that pattern to guide future automated edits. Nothing generic, nothing borrowed from someone else's presets, just your own decisions applied at scale.
You can re-run the calibration any time your style changes, and every AI-driven edit still lands in imagic's normal edit history, so you can review or undo it like anything else.
This isn't the full feature comparison, just the part that matters if you want an AI agent doing the work.
The specifics people ask before connecting an AI agent to their photo library.
MCP (Model Context Protocol) is an open standard that lets AI assistants connect to external tools. imagic's MCP server exposes its scan, cull, edit, and export tools so an AI agent can call them directly, the same actions you'd trigger by hand in the app.
Yes. imagic's MCP server works with any MCP-compatible client, including Claude Code, Claude Desktop, Cursor, Codex, and GitHub Copilot. Setup is a one-time config step per client, shown above.
No photos are uploaded anywhere. The MCP server runs locally alongside imagic and talks to your AI client over your own machine. Processing stays local, the same as using imagic directly.
No. You paste one setup command or config snippet once, then talk to your AI assistant in plain English. It calls imagic's tools for you.
No. Edits are stored as non-destructive adjustments in imagic's edit history, the same as manual edits. Nothing overwrites your original RAW or source files unless you explicitly export.
MCP and the CLI are part of imagic Max, a one-time €99 license with no subscription. The free 7-day trial includes full Max access, no card required, so you can try MCP and the CLI before deciding. They aren't included in Lite or Plus.
Try the full imagic Max workflow, MCP and CLI included, free for 7 days. No card required, no subscription, and nothing to cancel.