lightcraft/docs/mcp.md

LightCraft MCP server

lightcraft-cli mcp exposes LightCraft to AI agents through the Model Context Protocol: newline-delimited JSON-RPC 2.0 over stdio (protocol revision 2025-06-18; 2025-03-26 and 2024-11-05 clients are accepted). The server lives in crates/mcp (lightcraft-mcp, layer L5) and is hand-written: no async runtime, no C dependencies.

It runs in one of two modes:

Mode	Command	What it drives
Headless (default)	lightcraft-cli mcp [--demo] [FILES/FOLDERS…]	An in-process engine Session. Develop, render and export without a window.
Connect	lightcraft-cli mcp --connect [127.0.0.1:7980]	A running desktop app started with lightcraft --control 7980, through its loopback JSON-lines control channel (control-protocol.md). Adds the UI tools (screenshot, clicks, keys, pointer gestures).

Options: --library DIR opens (or creates) a persistent LightCraft library — the same crash-safe format the desktop app uses (~/Pictures/LightCraft Library by default there) — so ratings, edits and albums survive between sessions (with --demo, a new library is seeded with the demo photos). A library is open in one program at a time (catalog.lock in the library folder): while the desktop app has it open, --library on the same folder fails with "This library is already open in LightCraft (process N …)" — use connect mode to work with the running app instead; --demo starts the headless session with the procedurally generated demo library; --compact lists only the helper tools (see below). In connect mode the server starts even when the app is not running yet and connects on the first call (and reconnects if the app restarts).

Logs go to stderr; stdout carries only protocol messages.

Wiring it into a client

Build once: cargo build --release -p lightcraft-cli (binary: target/release/lightcraft-cli).

Claude Code

# headless, with a folder of photos imported at start
claude mcp add lightcraft -- /path/to/lightcraft/target/release/lightcraft-cli mcp ~/Pictures/shoot

# or: drive the running desktop app (start it with `lightcraft --control 7980`)
claude mcp add lightcraft-app -- /path/to/lightcraft/target/release/lightcraft-cli mcp --connect 127.0.0.1:7980

Or check a project-scoped .mcp.json into your repo:

{
  "mcpServers": {
    "lightcraft": {
      "command": "/path/to/lightcraft/target/release/lightcraft-cli",
      "args": ["mcp", "--connect", "127.0.0.1:7980"]
    }
  }
}

Codex

The macOS release includes a separate lightcraft-cli-<version>-macos-universal.zip; unpack it alongside the desktop app if you do not want to build from source. Use the CLI and app from the same release. Verify both downloads against that release's SHA256SUMS.txt.

Register the CLI's stdio server with Codex using its absolute installed path:

# Headless: a persistent library whose edits survive MCP restarts.
codex mcp add lightcraft -- "/absolute/path/lightcraft-cli" mcp --library "/absolute/path/library"

# Live desktop: start the app on the matching loopback port first.
"/absolute/path/LightCraft.app/Contents/MacOS/LightCraft" --library "/absolute/path/library" --control 7980
codex mcp add lightcraft-app -- "/absolute/path/lightcraft-cli" mcp --connect 127.0.0.1:7980

Choose one mode for a library: a headless server cannot open a library already locked by the desktop app. Connect mode edits the window's actual library and reconnects after the app restarts. An external library path keeps the catalog on that drive. Copy/move imports default to the library's Originals/ folder; add-mode imports keep the originals at their existing paths. The library path does not relocate settings, logs or optional models (see the README's settings/log locations). Make sure an external drive is mounted before starting either process.

codex mcp list confirms registration, not a successful tool call. After adding the server, restart the client's MCP connections and verify doc_inspect or query_photos against the expected library before editing. Configuration options are in the Codex MCP documentation.

From an installed release

The release packages ship lightcraft-cli alongside the desktop app, so no build is needed:

Install	CLI
Windows (MSI)	C:\Program Files\LightCraft\lightcraft-cli.exe by default (wherever you installed it otherwise), not on PATH
Linux (deb, rpm)	/usr/bin/lightcraft-cli
macOS	the separate lightcraft-cli-<version>-macos-<arch>.zip release asset (the .app holds only the desktop app)

# Windows, default install folder
claude mcp add lightcraft -- "C:\Program Files\LightCraft\lightcraft-cli.exe" mcp
# Linux, or macOS with the CLI unzipped onto PATH
claude mcp add lightcraft -- lightcraft-cli mcp

Other clients (Claude Desktop, Cursor, …)

Every stdio MCP client takes the same shape: a command plus args. For example claude_desktop_config.json:

{
  "mcpServers": {
    "lightcraft": {
      "command": "/path/to/lightcraft/target/release/lightcraft-cli",
      "args": ["mcp", "--demo"]
    }
  }
}

During development you can also point the client at cargo run --release -p lightcraft-cli -- mcp (with "cwd" set to the repository), at the cost of a slower start.

Shared core tools

These follow the same conventions as filmcraft #28. They are listed in both full and compact mode; existing documented helpers remain listed too. There are no hidden compatibility aliases.

Tool	Arguments	Result
command_list	filter?, enabled_only?	Command catalog
command_run	id, params?	Command result
command_batch	steps: [{id, params?}], stop_on_error?	Counts and per-step results; stops on error by default
doc_inspect	none	Library state and catalog counts
render_preview	id?, max_side?, format?	Image; neither changes selection nor saves a file

Each edit in a batch has its own undo step. Connect mode also lists ui_inspect and ui_screenshot (the latter has no output-path argument). The existing inspect_ui, screenshot, list_commands, run_command and render_photo keep their documented arguments. Ports and connect mode are unchanged.

Every listed tool has a title and all four MCP hints. Generated command tools are conservatively marked as edits. Helpers with optional output paths are marked as writers; export is not idempotent because the default conflict policy chooses a new filename on repeated calls. Unknown helper argument keys return JSON-RPC -32602 naming the key and accepted arguments. Generated cmd_* tools and nested command params remain free-form: the registry has prose parameter docs, not machine-readable schemas. Existing command validation (including strict export options) is preserved. Escaped tool panics return isError: true, and the session keeps serving; backend panics during resource reads or tool listing return an internal error. The headless backend owns its session directly, without a session mutex.

lightcraft://document and lightcraft://commands contain JSON matching doc_inspect and command_list. Existing resources remain available. Requests declaring MCP 2026-07-28 in per-request _meta receive resultType: complete and list/read cache hints; reads are not cached.

Tools

Helpers

Tool	Does
list_commands {filter?}	Every command: id, label, menu, shortcut, parameter doc, enabled now
run_command {command, params?}	Run any command by id
import {paths, album?, mode?, destination?, organize?, rename?}	Import files/folders (folders are scanned recursively); the first new photo becomes active. mode: "copy" copies into destination (default: the library's Originals/), mode: "move" moves there (each original and its XMP sidecars are removed from the source only after the copy is verified and catalogued; duplicates and failures keep their sources; the result's moved / kept list what moved and what stayed, and why; undo leaves the moved files at the destination), filed by organize: date (YYYY/YYYY-MM-DD), month, flat or a folder template such as {date:%Y}/{date:%Y%m%d} (→ 2026/20260114; capture date, else the import date; always inside the destination), and named by the rename template (run_command photo.renameTokens lists the tags)
query_photos {filter?, sort?, offset?, limit?}	Photos in the current view (or matching a catalog Filter)
select_photos {ids, active?, mode?}	Set the selection / active photo. Every id (and active) must be in the library: an unknown id is a tool error (no such photo 9999) and the selection and active photo are left as they were
list_controls {section?}	Every develop slider: id (light.exposure…), range, default, current value
get_develop {id?}	Full develop-settings JSON, with process: the rendering process the photo is on (process-versions.md)
set_develop {id?, values?, settings?, label?}	values: {controlId: number}; settings: partial develop JSON deep-merged. Undoable. A crop.angle value turns the crop as crop.straighten does: kept if it still fits, else shrunk to the largest crop of its aspect inside the photo (settings is merged as given, unfitted)
apply_preset {preset, amount?, ids?}	Apply a preset (ids from cmd_presets_list)
crop {id?, rect?, angle?, reset?}	Normalized crop rect [x0,y0,x1,y1] and straighten angle; at least one of rect, angle, reset: true. The crop always ends inside the photo: an angle alone keeps the crop if it still fits, else shrinks it to the largest of its aspect that does
render_photo {id?, size?, format?, path?}	Render with current settings → image content (PNG, or JPEG with format: "jpeg"), long edge size (default 1024)
export {path | dir, id? | ids?, format?, longEdge? | shortEdge? | width?/height? | megapixels? | percent?, dontEnlarge?, ppi?, quality?, colorSpace?, bitDepth?, …}	Full-quality render to .png / .jpg / .tif / .webp / .avif; format: "original" copies the file + an XMP sidecar with the edits, format: "dng" writes raw photos as DNG with the edits embedded. No size param = 3000 px long edge; longEdge: 0 = full size (cropped, native resolution); width + height fit either orientation; dontEnlarge defaults to true

Tools taking id make that photo active first; without it they act on the active photo.

Connect mode only

Tool	Does
screenshot {maxSize?, format?, path?}	The app window as an image, after pending renders finish
inspect_ui	View, panel, window/image rects, selection, status
set_ui {state}	Merge UI state, e.g. {"view": "detail"}
list_widgets {filter?} / click {widget | x,y, count?, button?}	Widgets by automation id; real egui clicks (button: "right" right-clicks)
press_key {key, cmd?, shift?, alt?} / type_text {text}	Keyboard input (shortcuts)
clipboard {action, text?}	Cut, copy or paste in the focused text field (⌘X / ⌘C / ⌘V; refused when none has the focus; cut and copy write the system clipboard); inspect_ui → copied is what was copied
pointer_gesture {events}	Gestures in normalized image coordinates (brush strokes, gradients, crop handles)

In headless mode these return a tool error explaining how to start the app.

One tool per command

Everything is a command in LightCraft, so tools/list also contains one tool per entry of the command registry (engine commands, plus the app's UI commands such as view.detail when connected): the id with . replaced by _ and a cmd_ prefix — photo.rate → cmd_photo_rate, develop.set → cmd_develop_set, edit.undo → cmd_edit_undo. Arguments are the command's JSON params (documented in each tool's description and by list_commands). Pass --compact to leave these per-command tools out when a client struggles with a large tool list; run_command still reaches every command.

Resources

resources/list / resources/read serve JSON snapshots:

URI	Content
lightcraft://library	Source, filter, sort, selection, undo/redo labels (library.state)
lightcraft://photos	Photos in the current view (catalog.query)
lightcraft://photo/active	Everything about the active photo (photo.inspect)
lightcraft://develop/active	The active photo's develop settings (develop.get)
lightcraft://controls	Every develop control with its current value (develop.controls)

Example session

→ {"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"me","version":"1"}}}
← {"jsonrpc":"2.0","id":1,"result":{"protocolVersion":"2025-06-18","capabilities":{"tools":{…},"resources":{…}},"serverInfo":{"name":"lightcraft",…},"instructions":"…"}}
→ {"jsonrpc":"2.0","method":"notifications/initialized"}
→ {"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"set_develop","arguments":{"values":{"light.exposure":0.7}}}}
← {"jsonrpc":"2.0","id":2,"result":{"content":[{"type":"text","text":"{…}"}],"isError":false,"structuredContent":{"ok":true,"controls":[{"id":"light.exposure","value":0.7}]}}}
→ {"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"render_photo","arguments":{"size":768}}}
← {"jsonrpc":"2.0","id":3,"result":{"content":[{"type":"image","data":"iVBORw0…","mimeType":"image/png"},{"type":"text","text":"{\"id\":21,\"width\":512,\"height\":768}"}],"isError":false}}

Errors from commands (unknown control, nothing selected, app not reachable) come back as tool results with isError: true so the model can read and correct them; malformed JSON-RPC gets the standard error codes (-32700 parse, -32600 invalid request, -32601 method not found, -32602 invalid params, -32002 resource not found).

With --library, a command whose change can't be written to disk (disk full, …) is an error too: saved in memory but not written to disk: <reason>; LightCraft will retry — the change is applied in the session and written by the next successful save (see control-protocol.md).

One-shot commands: lightcraft-cli run

For agents that prefer a shell over an MCP session: run any chain of commands in one process and read one JSON line per command ({"command", "ok", "result" | "error", "ms"}; non-zero exit status on failure). A word without = starts the next command; key=value values are JSON when they parse, else strings; a '{…}' argument merges a JSON object into the params. Quote values with brackets or spaces for your shell ('ids=[3,4]'; zsh treats […] as a glob).

# headless: import, edit, export
lightcraft-cli run --import ~/Pictures/a.dng develop.set control=light.exposure value=0.7 \
    develop.auto app.export path=/tmp/a.jpg shortEdge=1080 colorSpace=displayP3
# a persistent library: edits are saved, later invocations see them
lightcraft-cli run --library ~/lc-lib --import ~/Pictures/shoot library.info
lightcraft-cli run --library ~/lc-lib library.select ids=[3] develop.get
# the running app (same commands, plus ui.* methods)
lightcraft-cli run --connect ui.set view=detail ui.screenshot path=/tmp/ui.png
# JSON lines from a file or stdin: {"command": id, "params": {…}} or {"method": "ui.inspect"}
lightcraft-cli run --demo --script steps.jsonl --keep-going

Other CLI subcommands

lightcraft-cli render in.dng -o out.tif --opt colorSpace=displayP3 --opt bitDepth=16 --opt percent=50
lightcraft-cli render in.dng -o out.jpg --set light.exposure=0.5 --set light.contrast=20 --size 2048
lightcraft-cli render in.jpg -o out.png --settings look.json --preset <presetId>
lightcraft-cli commands [--json]   # the command registry
lightcraft-cli controls [--json]   # develop control ids and ranges
lightcraft-cli calibrate --max 300 ~/Pictures/2026   # camera colour profiles (docs/camera-preview-colour.md)

Tests

• crates/mcp/tests/e2e.rs — M0.9 acceptance: over the stdio framing, set exposure and render; checks the decoded PNG gets brighter/darker. Runs headless and through the TCP transport (Remote) against a stand-in control server; also import → render → JPEG export of a real file.
• apps/lightcraft-cli/tests/cli.rs — spawns lightcraft-cli mcp with real pipes; render; commands.

Export progress and cancellation

Headless direct export, command_run / run_command with app.export, and cmd_app_export calls report photo-count progress when params._meta.progressToken is a string or number. Notifications are strictly increasing, at most ten per second plus the final total. No token means no notifications. ping is answered at photo boundaries; other requests wait in order until the export ends. EOF lets a pending export finish and preserves queued requests.

notifications/cancelled with params.requestId stops the matching export before its next photo, suppressing its response. Unknown/completed request ids are ignored. A photo already being processed finishes first: completed photos remain, each written with the existing atomic file writer, and there are no partial files to delete. Unrelated outputs and earlier exports are untouched. A failed export is an isError tool result. Transport mutexes recover poisoning.

Connect mode and exports inside command_batch remain synchronous and do not report MCP progress or cancellation. Use a direct headless export call for this behavior.