AI tool
Refs
Refs is a reference library of 2,449 launch films broken down into shot-by-shot blueprints and prompts. It helps founders and directors plan product films with Claude by searching for relevant references, pulling a blueprint, and planning the film...

About Refs
Refs is a library of over 3,000 films, each deconstructed into a 15-plate blueprint with a corresponding recipe prompt. It's built to give Claude and other coding agents specific, measured references for planning video projects. The tool plugs into Claude Code, Codex, and Cursor as an MCP, letting the AI search the library and pull a blueprint to plan a film shot by shot.
Review
Refs tackles a specific problem with AI-generated video planning: the tendency to produce generic, averaged-out results. The approach is to supply a large, hand-measured database of existing films that a user can point to as a concrete direction. It doesn't generate video itself; it functions as a reference layer that sits between a user's request and the AI's planning stage.
Key Features
- Measured film blueprints: Each of the 3,000+ films is broken down cut by cut into a 15-plate structure covering hook, beats, shots, type, color, sound, and script. The library currently holds measurements for over 60,000 individual shots.
- Recipe prompts: Every film comes with a ready-made prompt that can be handed directly to Claude to guide its output toward that specific reference's style and structure.
- Formulas and recommendations: The tool identifies common patterns across strong films within a format and surfaces which references are most applicable for SaaS, app, or AI tool launch films.
- MCP integration: Refs connects to Claude Code, Codex, and Cursor as an MCP. When asked for a launch film, Claude can search Refs, retrieve a blueprint, and plan the film shot for shot based on that data.
Pricing and Value
Browsing the film library is free. The Pro tier costs $16 per month and unlocks access to every plate, every recipe prompt, the formulas, and the MCP integration. An annual billing option with a 50% discount is listed on the launch page.
Pros
- Supplies a concrete reference system that replaces vague prompting with measured, shot-by-shot data from real films.
- The library includes launch films from companies like Linear, Runway, Vercel, OpenAI, and Perplexity, which makes the references directly relevant to tech product videos.
- MCP integration means the reference lookup and blueprint retrieval happen within the coding agent's workflow, not in a separate tool.
- Free browsing lets users explore the full library before committing to a paid plan.
Cons
- The tool only plans and provides references; it won't generate or edit video files. Users still need separate tools and skills to produce the actual film.
- Claude works from text descriptions of shots and edits rather than the visual or temporal data of the original film, which limits the fidelity of the reference signal.
- Refs is not well suited for creators who work outside the supported coding agents (Claude Code, Codex, Cursor) or who don't use an AI-assisted planning workflow.
Refs fits a narrow but defined gap for founders and teams who already use Claude or similar tools to plan video projects and want to move past generic output. It's most useful when you have a clear sense of a reference film's feel and need a structured way to hand that direction to an AI. The tool won't help someone who needs finished video files or who prefers to direct a project without an AI intermediary.









