New in 1.3 · MCP + skills

Talk to your footage from your AI agent

Reelary gives Claude, Cursor, Codex and other agents a tool they never had: search the videos and photos already on your computer by what is in them, then cut the matches into a video. The app does the heavy lifting on your machine; the agent just asks.

Install in one line

$ npx reelary

This downloads the signed Reelary app (or keeps the one you have), starts it in the background with a tray icon and no window, connects it to Claude Code, Claude Desktop, Cursor and Codex, and adds the Reelary skills. Restart your agent and ask about your footage. Requires Node.js 18 or later.

Prefer to pick the pieces yourself:

# skills only, for any of 60+ agents
$ npx skills add reelary/skills

# Claude Code plugin: skills + MCP server
/plugin marketplace add reelary/skills
/plugin install reelary@reelary

Without the app, the skills offer to install it the first time you ask about your footage. Source: reelary/skills (skills and plugin manifests) and reelary/cli (the reelary npm package and MCP server).

What you can ask

• "Index ~/Movies and find drone shots of the coast."
• "Find b-roll from my footage for this voiceover: …"
• "Find where she talks about the budget in the interview."
• "Cut a 30-second 9:16 recap of my Lisbon trip."
• "Rough cut this script from my footage and export it for DaVinci Resolve."

The agent finds shots with exact start and end times, lines up one clip per sentence of your script with alternatives, saves the result as a Reelary project you can open and trim in the editor, and exports an MP4 or a timeline for Premiere Pro, Final Cut Pro or DaVinci Resolve. A walk-through: a rough cut from a script, with Claude Code.

How it works

The Reelary app keeps a local API open on 127.0.0.1 while it runs. Your agent starts npx reelary mcp, a small MCP server that translates tool calls into requests to that API, and launches Reelary in the background if it is not running. Everything that touches your files - indexing, the vision model, search, rendering - happens in the app, on your machine, exactly as when you click through it yourself.

Long jobs such as indexing a drive or rendering an export run in the background; the agent checks their progress instead of waiting on one long call. The skills tell the agent how to use the tools well: when to search versus rough-cut, how to pick between alternatives, and which features need Pro.

What the agent sees, and what it does not

Your video never leaves the computer. The agent gets text back: file paths, time ranges, a one-line description of each shot and, with Pro, the line of dialogue that matched. It can ask for up to eight small preview frames per search when it needs to look. Whatever the agent receives is sent to its model provider, the same as anything you type into the chat. If that is a problem for a client project, use an agent running a local model, or use the Reelary app directly - it never sends anything. More on this in what an AI agent actually sees of your footage.

Skills

Skill	For
reelary	Start here: installs and connects Reelary, indexes folders and drives, fixes common errors.
find-footage	Shots by description, photos, quotes, people and places (the last three with Pro).
broll-finder	Cutaways for each line of a voiceover, interview or podcast.
rough-cut	Script, outline or shot list to an editable cut; montages and highlight reels.
vertical-video	Shorts, Reels and TikToks in 9:16 or 1:1, reframed from 16:9 footage.
nle-export	FCPXML, XML or EDL for Premiere Pro, Final Cut Pro and DaVinci Resolve. (Pro)

MCP tools

Tool	Does
reelary_status	App version, license and library size.
reelary_list_folders	Folders in the library.
reelary_add_folder	Add and index a folder or drive (free, unlimited).
reelary_remove_folder	Drop a folder from the index; files on disk are untouched.
reelary_search	Search by meaning; returns file paths with exact start and end seconds.
reelary_rough_cut	One scene per script line, best clip plus alternatives.
reelary_save_project	Save the draft as a project you can open in the editor.
reelary_open_in_app	Bring up the Reelary window on a project.
reelary_export_video	Render an MP4 (free tier limits apply).
reelary_export_nle	FCPXML or EDL that references your original files. Pro.
reelary_job_status	Progress of indexing and export jobs.
reelary_list_people	People recognised in the library. Pro.
reelary_person_clips	Every clip a person appears in. Pro.
reelary_list_places	Where footage was shot, from phone, drone and GoPro GPS. Pro.

Or from the terminal

The same package works as a plain command line, handy for scripts:

$ npx reelary status
$ npx reelary index ~/Movies
$ npx reelary search "sunset over water" --json
$ npx reelary cut "Sunrise over the city. Coffee being poured." --aspect 9:16 --save Morning
$ npx reelary export <project_id> --nle fcpxml

Requirements

• Reelary 1.3 or later, direct download from reelary.app (Mac or Windows).
• Node.js 18.17 or later, for npx.
• An agent that supports MCP or skills.

Common questions