Jockey is TwelveLabs' AI agent that organizes, searches, and assembles content across your entire video library using Pegasus and Marengo for intelligent video understanding and retrieval.

How does Jockey search videos?

Jockey uses natural language queries to search video content, powered by Marengo for retrieval and Pegasus for understanding, enabling precise semantic searches.

Can Jockey assemble clips from multiple videos?

Yes, Jockey can intelligently assemble relevant clips from across your video library into a coherent final video based on your instructions.

What video formats does Jockey support?

Jockey supports common video formats like MP4, MOV, and AVI, though specific formats may vary; check TwelveLabs documentation for full details.

Is Jockey suitable for large video libraries?

Yes, Jockey is designed to scale, efficiently indexing and searching thousands of hours of video content using AI-powered indexing.

Does Jockey require technical expertise to use?

No, Jockey is designed with a user-friendly interface for non-technical users, while also offering an API for developers to integrate into workflows.

Jockey - AI AI Search tools - Free trial, pricing intro, performance review, official site access and online experience

What is Jockey?

Jockey is an AI agent developed by TwelveLabs that organizes, searches, and assembles content across your entire video library. It is built on Pegasus and Marengo for intelligent video understanding and retrieval. Users can ask anything about their footage, find specific clips using plain language, and generate highlight reels automatically. It acts as a "agent layer for video understanding," enabling teams to interact with video archives conversationally.

Application scenarios

Marketing teams
Automatically tag every ad by hook, talent, format, and tone as soon as it lands, then compare performance across formats.
AI-native creators
Talk to your entire footage archive through an assistant like Claude (ChatGPT integration coming soon), search by describing a moment, or find any quote you've said on camera.
Developers
Use video understanding as an API to search, extract, and embed what's inside users' footage, with multimodal embeddings and natural-language search.
Ad performance analysis
Detect winning hooks (the first three seconds that hold attention) across your whole library.
Talent and logo recognition
Find every ad a creator, spokesperson, or brand mark appears in without opening a file.
Highlight reel creation
Point Jockey at a topic and get an edit-ready cut of your best moments back.

Core Features

Automatic Ad Tagging
Every ad is watched and labeled by hook, talent, format, and tone the moment it lands.
Winning Hook Detection
See which opening three seconds actually hold attention across your entire library.
Talent & Logo Recognition
Find every ad a creator, spokesperson, or brand mark appears in without opening a file.
Format & Trend Breakdown
Compare performance across formats and spot patterns before your next shoot.
Connect to Claude
Plug your archive into the assistant you already use and talk to your footage; ChatGPT integration is coming soon.
Search How You Think
Describe a moment in plain language and Jockey finds the exact clip, not the folder.
Find Any Quote
Search everything you've ever said on camera and jump straight to the line.
Instant Highlight Reels
Point Jockey at a topic and get an edit-ready cut of your best moments back.
Multimodal Embeddings (for developers)
One embedding across video, image, audio, and text to power your own retrieval and ranking.
Natural-Language Search (for developers)
Ship a search bar that understands what's in your users' videos, not just the filenames.
Structured Data Extraction (for developers)
Define a JSON schema and get back timestamped, machine-readable metadata for every video.
Frame-Accurate Timestamps (for developers)
Every result maps back to the exact moment, so you can deep-link straight into playback.

Target users

Jockey is designed for marketing teams who need to analyze ad creative libraries, AI-native creators who want to search and remix their footage archives, and developers who need to integrate video understanding into their own applications via API. It suits anyone who works with large video collections and needs intelligent, natural-language-driven access to their content.

How to use Jockey?

To get started, visit the Jockey website and sign up for limited early access. Once accepted, you can upload your video library or connect your existing archive. You can then use natural language to ask questions, search for clips, or generate highlight reels. For developers, the platform offers API access for multimodal embeddings, natural-language search, structured data extraction, and frame-accurate timestamps. The site offers a "Try for free" option and a demo to explore capabilities.

Pricing and free trial

The website text does not provide specific pricing details. It mentions "Research Preview" and "currently accepting limited sign-ups," with a "Try for free" option available. No paid tiers or subscription costs are stated.

Effect review

Jockey's feature set is impressive for a research preview, especially the ability to tag ads automatically and detect winning hooks in seconds. The integration with Claude (and soon ChatGPT) makes it practical for creators who already use AI assistants. For developers, the multimodal embeddings and frame-accurate timestamps offer real utility for building custom video search tools. However, since it's in limited preview, the actual performance and scalability are not yet verified by broad user feedback. Overall, Jockey positions itself as a powerful agent for video understanding, but its real-world effectiveness will depend on how well it handles diverse, large-scale video libraries.

Jockey