Refs icon

Refs

Refs is a video reference library for AI agents and creators, covering launch films, motion pieces, explainers, and other short-form video formats. It measures source videos and turns them into cut maps, blueprints, formulas, recipes, and MCP-accessible planning inputs.

Refs

A measured reference library for AI video work

Refs is a video reference library designed for AI agents and the people directing them. It catalogs films such as launch videos, motion-design pieces, and explainers, then analyzes them as production references rather than simply presenting them as a mood board.

The service turns selected videos into measured cut maps, structured blueprint plates, cross-film formulas, and copy-ready recipes. Through its MCP connection, an agent can search the archive, retrieve reference analysis, and plan a new video from a brief and one to five references while keeping the original source footage on X or YouTube.

Video reference analysis for agents

Searchable video archive

Browse a library of 3,033 films, including launch films, motion pieces, explainers, and other brand and product videos, with category and keyword-oriented archive browsing.

Measured cut maps

Review frame-by-frame shot boundaries, cut timelines, duration, shot counts, average shot length, cuts per minute, and selected audio or platform observations.

Multi-plate blueprints

Open structured blueprints that describe a film’s format, hook, beats, shots, visual style, sound, script, production plan, edit recipe, and QA considerations.

Cross-film formulas

Compare films of the same format through formulas that identify recurring structures and provide template prompts for adapting a method to a new concept.

Runnable recipes

Use copy-ready recipes and prompts to turn a reference into an original production starting point, with the source film’s method rather than its footage as the model.

MCP agent access

Connect the archive to Claude Code, Codex, or Cursor through a streamable HTTP MCP server and use tools for search, blueprint retrieval, formula analysis, and video planning.

Practical ways to use Refs

  • Research a video format

    Search for films in a particular format, such as a founder launch or short product reveal, then study their timing, shot grammar, and structure before drafting a concept.

  • Turn a reference into a brief

    Use a blueprint to translate a successful reference into a production outline covering beats, shots, visual treatment, sound, script, and edit decisions.

  • Plan a new film with an agent

    Give an agent a product brief and a small set of references so it can produce a beat sheet, shot list, and build prompt grounded in existing examples.

  • Develop a repeatable format

    Compare several films in the same format to identify shared patterns and adapt the resulting formula without copying the original footage.

  • Add references to a coding workspace

    Connect Refs through MCP when working in Claude Code, Codex, or Cursor so reference search and analysis are available inside the agent workflow.

Pros and Cons

Pros

  • Provides concrete production analysis instead of only visual inspiration, including shot timing, pacing, structure, and sound observations.
  • Combines individual film breakdowns with formulas distilled across multiple films of the same format.
  • Offers an agent workflow through MCP for Claude Code, Codex, and Cursor.
  • Free access supports archive browsing, source playback, measurements, and initial blueprint plates.

Cons

  • The archive’s deeper outputs, including complete blueprints, recipes, formulas, and MCP access, require Pro.
  • Refs analyzes and links to source videos rather than hosting or rehosting the footage.
  • Credit-based usage applies when taking content out through blueprint, formula, or planning operations.

FAQ

What is Refs?

Refs is a video reference library for agents. It lets users search films, inspect measured cut maps, retrieve blueprints and formulas, and use references to plan new videos.

What is available for free, and what requires Pro?

Browsing the archive, playing source videos, viewing measurements, and reading the first blueprint plates are available free. Pro unlocks complete blueprints, recipes, formulas, and MCP access.

How do I connect Refs to an AI coding client?

Create a Refs key in the account page, store it as REFS_KEY, then connect the MCP server at https://refs.video/mcp using a Bearer token. The documented clients are Claude Code, Codex, and Cursor.

What can the Refs MCP do?

The MCP tools can search videos, retrieve a specimen's blueprint, list or retrieve formulas, and plan a video from a brief and one to five references. Searching is free; other operations use credits.

Does Refs host the source videos?

Refs does not rehost the videos. Each film plays from its original post on X or YouTube, while Refs provides its measurements, analysis, blueprints, formulas, and recipes.

Quick Facts

Category
Video reference library
Primary users
AI-assisted video creators and agents
Archive size
3,033 films and 67,703 measured shots
Source media
Original posts on X or YouTube; Refs does not rehost footage
MCP clients
Claude Code, Codex, and Cursor
Access model
Free browsing with Pro features and credit-based extraction

Alternativas a Refs

Hype icon

Hype

Hype is a web tool for finding trending YouTube topics by category, time range, and scoring mode. It helps creators spot emerging ideas, inspect source videos, and decide what to cover next.

Paperpal icon

Paperpal

Paperpal 是一款面向学术写作的 AI 工具,提供文献阅读、语言润色、引用、查重、AI 检测和投稿前检查。它适合需要在英文论文和学术材料上进行写作、修改与合规自检的学生、研究人员和教师。

VForms icon

VForms

VForms lets you place questionnaires directly inside YouTube videos so viewers answer in context instead of switching to a separate form. It is aimed at collecting feedback and insights through timestamped, embeddable video forms.

Scite icon

Scite

Scite es una plataforma de investigación con IA para descubrir artículos, verificar el contexto de citas y obtener respuestas basadas en literatura académica. Ayuda a evaluar evidencia en fuentes de texto completo, patentes y materiales relacionados.

Radar icon

Radar

Radar is a self-hostable media intelligence and social listening app for monitoring public sources, analyzing mentions, and generating briefs or content outputs with optional Claude AI. It is aimed at individuals and small teams that want an enterprise-style workflow without a cloud account.

SHODH icon

SHODH

Shodh is an AI-powered research and strategy platform for business questions that need structured investigation, source-backed evidence, and auditable recommendations. It is aimed at founders, operators, consultants, investors, and corporate development teams.

Refs - AI Tool, Features, Use Cases & Alternatives | UStack