Lightroom AI Bridge 1.0.0

Lightroom AI Bridge lets Claude edit your photos in a real, running Adobe Lightroom Classic. It moves the actual Develop sliders, looks at a render of the result, and adjusts again, so every change is a normal non-destructive step in your History panel. Your RAW files are never modified, and Lightroom's generative tools are only used when you ask for them.

This is the first public release.

Get started

1. Download the source zip below (or git clone), then add ClaudeBridge.lrplugin in Lightroom's Plug-in Manager.
2. Select a photo and run ./lrc ping to confirm the connection.
3. Run ./skills/install.sh and type /lightroom-bridge-pro in Claude Code, or connect any MCP client.

Step-by-step walkthrough: user guide. Every command: command reference.

Requirements: macOS, Adobe Lightroom Classic 15.3 or later, Python 3 (already on macOS). The MCP server additionally needs Python 3.10+, provisioned automatically by uv.

Added

• Lightroom plug-in (shown in Lightroom as "Claude Bridge", internal plug-in
  version 1.2.0): exposes Lightroom Classic's Develop module over two localhost
  sockets, built only on Adobe's public SDK. Inbound commands are parsed as data
  in an empty sandbox, never executed as code.
• lrc command line tool with 40+ commands: tone, white balance, crop and
  straighten, AI and geometric masks, presets (list, apply, save as a real
  Develop preset), snapshots, undo and redo, ratings, flags and labels with
  verified per-photo writeback, AI Denoise and Remove People on supported
  formats, and full-fidelity render previews for before/after comparison.
  Pure Python 3 standard library.
• MCP server (lightroom-bridge) exposing the same operations as 46 tools
  for any MCP client, with previews returned as inline images and seven kb_*
  tools for the knowledge base. Shares one background daemon with the CLI.
• Claude Code skills: lightroom-bridge for everyday editing and culling,
  and lightroom-bridge-pro for a full professional edit of the selected photo
  with no questions asked. Installer: skills/install.sh.
• Knowledge base scaffold (kb/): the recall, apply, learn loop with its
  card format, consolidation and revision rules, and bridge/kb.py tooling.
  Ships empty by design, so each photographer's taste profile is built from
  their own corrections.
• Helper scripts in bridge/: contact-sheet folder review, the two-stage
  culling funnel with verified writeback, fast bulk metadata writeback,
  verified per-photo batch apply, full-JSON settings apply, tilt and plumb
  measurement, and an attention-hierarchy check.
• Documentation: a step-by-step user guide, a full command reference, and
  agent cheat-sheets (CLAUDE.md, AGENTS.md).

Known limitations

• macOS only. Requires Lightroom Classic 15.3 or later.
• Geometric masks (radial, linear) cannot be positioned through Lightroom's SDK;
  they land at Lightroom's default spot. AI masks are unaffected.
• Mask values cannot be read back through the SDK; masks are verified by render.
• Not yet published to PyPI; install from source (see the user guide).