FrameThrower — Film Still Library with a REST API, MCP Server & SDK

What

FrameThrower is a searchable cinematic film still library and visual reference platform designed for filmmakers, directors, cinematographers, and visual artists. It replaces manual reference gathering and generic stock photo searches by indexing film stills that can be retrieved via natural language, technical cinematography attributes, and visual similarity. The platform is positioned both as a collaborative creative tool for pre-production pitching and as an infrastructure layer accessible via a REST API and Model Context Protocol (MCP) server.

Features

• Semantic and Cinematographic Search: Enables users to find film frames by querying natural language, visual similarity, or specific parameters such as lighting, lenses, color, and camera angles.
• Script-Driven Lookbook Builder: Converts raw screenplay text into scene-by-scene pitch boards populated with real cinematic frames to expedite pre-visualization.
• Collaborative Moodboard Canvas: Allows users to drag and drop reference stills onto an organizational canvas categorized by scene, sequence, or look, shareable via a single URL.
• Developer REST API and SDK: Provides an 11-endpoint programmatic interface and npm package (framethrower-ai) with pay-per-usage billing to integrate cinematic search into tools like ComfyUI and custom pipelines.
• Model Context Protocol (MCP) Server: Exposes four mid-conversation agent tools (search, craft, similar, and frame detail) over OAuth 2.1 to support AI agent retrieval workflows.

Helpful Tips

• Account for Stills Usage Restrictions: Stills in the library are provided in reduced resolution solely for educational, reference, and commentary purposes, meaning they cannot be repurposed as production-ready visual assets.
• Test the MCP Integration for Pre-Visualization: Teams building agentic media workflows can use the provided MCP server endpoint to let conversational AI assistants fetch visual citations during script breakdown tasks.
• Monitor API and MCP Usage: Because developer and agent access operates on a pay-per-usage model past the initial free credit tier, teams should establish query limits to manage pipeline costs effectively.

OpenClaw Skills

The documentation contains no native integration with OpenClaw. As a hypothetical use case, an OpenClaw automation agent could interface with FrameThrower’s REST API or MCP endpoints to power a visual pre-production skill that parses incoming script treatments and automatically retrieves matching reference stills for directors and storyboard artists. If implemented, such a workflow could streamline visual development across production teams by programmatically connecting script analysis with curated cinematic references.