Twelve Labs launches Jockey video AI agent

// 1d agoPRODUCT LAUNCH

Twelve Labs launches Jockey video AI agent

Jockey is a conversational video intelligence agent by Twelve Labs designed to search, analyze, and reason across entire media libraries. Currently in research preview, the agent can plan multi-step video workflows, edit clips, and integrate with LLMs like Claude via the Model Context Protocol.

// ANALYSIS

Jockey represents a significant evolution from passive search to active agentic video manipulation, but its ultimate utility depends heavily on processing cost, query speed, and integration friction for large media libraries.

* Native video-first reasoning bypasses fragile and expensive transcription/tagging middleware.

* Supporting Model Context Protocol (MCP) allows seamless ecosystem integration with popular LLM assistants out-of-the-box.

* Latency and API cost scaling remain the primary hurdles for deploying agentic video search across massive enterprise archives.

* The transition from an open-source demo to a research preview signals a push towards commercializing custom video agent infrastructure.

// TAGS

jockeyartificial-intelligencevideo-searchmcptwelve-labsagentproductivitytool-use

DISCOVERED

1d ago

2026-07-21

PUBLISHED

1d ago

2026-07-21

RELEVANCE

8/ 10

AUTHOR

[REDACTED]

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

NEWS30m ago

Google allocates massive compute to Gemini 4

Google CEO Sundar Pichai announced that the company is allocating substantial compute capacity to build Gemini 4, a significantly larger foundation model designed to push the boundaries of frontier AI. The move underlines Google's commitment to scaling its AI infrastructure to maintain leadership in state-of-the-art AI development and performance.

MODEL32m ago

Researchers unveil OMG-VLM for multimodal graph processing

OMG-VLM is a newly unveiled open-source vision-language model designed specifically for processing multimodal graphs containing text and image elements. By making the model open source, researchers aim to enhance multimodal data analysis and facilitate advanced visual-textual graph processing across various research and domain applications.

UPDATE46m ago

Saravia Builds DAIR.AI Interface via Fable 5, GPT-5.6

Elvis Saravia (@omarsar0) demonstrated a multi-model workflow for building a new DAIR.AI community interface. He brainstormed concept designs with Fable 5 to produce an HTML artifact, which was then passed to GPT-5.6-Sol to construct the final interface.