Skip to content
Jockeyby TwelveLabs
Research Preview

The agent layer for
video understanding

Try for free

Currently accepting limited sign-ups. Register now.

Ask anything

Marketing Teams

Know every ad you've ever run.

Get started

Automatic Ad Tagging

Every ad watched and labeled by hook, talent, format, and tone the moment it lands.

AI-Native Creators

Your whole archive, on tap.

Get started

Connect to Claude

Plug your archive into the assistant you already use and talk to. ChatGPT coming soon.

Developers

Video understanding, as an API.

Get started

Multimodal Embeddings

One embedding across video, image, audio, and text to power your own retrieval and ranking.

Integration

Ask your library from the chat you already use.

Connect your photo and video library, then ask for a person, place, scene, or moment directly inside Claude. No new interface to learn. ChatGPT coming soon.

Find all the hikes with mountain views in my library.

Here are the hikes with mountain views from your library. Bookmark any that look good, or tell me more about what you’re looking for, like length, difficulty, or scenery, and I can refine the list.

Cascade Canyon Trail

Grand Teton National Park

Skyline Ridge Trail

Smith Wilderness Reserve

Cloudline Overlook

Fernwood Ridge Reserve

Jockey

Share

Reply to Claude

Sonnet 4.5
API & SDK

Bring Jockey into your own product.

Get an API keySDK Reference
Jockey
from twelvelabs import TwelveLabs, ResponseInputItem
client = TwelveLabs(api_key="<YOUR_API_KEY>")
response = client.responses.create(
model="jockey1.0",
knowledge_store_id="<YOUR_KNOWLEDGE_STORE_ID>",
session_id="<YOUR_SESSION_ID>",
input=[
ResponseInputItem(
type="message",
role="user",
content="Now list the three most important moments as bullet points.",
)
],
)
The models

Built on the models that understand video.

Marengo finds it, Pegasus describes it, Jockey turns it into something you can query.

Marengo

Indexes media by meaning.

Makes video and photo libraries searchable by phrase, image, clip, concept, action, object, place, or person.

In production

Teams are already querying their video.

Marketers, creators, and builders running libraries through Jockey from ad tagging to streamlined creative workflows.

Jockey moves us from ‘AI writes a description a human has to read’ to ‘AI emits structured data we can query.’ That’s the jump we needed.
Jannek Zechner
CTO · Human
It finds the exact moment I need across hours of footage in seconds. It has genuinely tightened my edits and saved me real time in post.
Kyle Chalmers
Content Creator · KC Labs
Jockey understands what happens inside a video, not just the labels on a segment. For sports analytics, that gap is the whole product.
Daniel Opata
Founder · LensX
Pricing

Start free. Scale as you go

Pick a tier by how much media you store. Every plan runs the full Marengo + Pegasus engine.

Free

Get started and explore the basics

$0/ mo
  • 5GB knowledge store
  • Basic rate limits

Plus

More room and higher limits for steady projects

$20/ mo
Monthly−17%
  • 100GB knowledge store
  • Higher rate limits

Pro

Recommended

Maximum headroom for heavy, large-scale work

$100/ mo
Monthly−17%
  • 500GB knowledge store
  • Highest rate limits

Plans renew automatically. Cancel anytime — you keep access until the end of your billing cycle.

FAQ

Common questions

Yes. When you delete your library or account, we permanently remove your original files, the generated index, associated metadata, and your usage history. You can manage deletions directly from your account settings.

Jockey supports standard video (MP4, MOV, AVI, MKV, WebM) and photo (JPEG, PNG, HEIC, HEIF, WebP, BMP, GIF) formats. Please refer to the documentation for specific storage and per-file size limits.

Jockey uses multimodal AI, analyzing visual, motion, and contextual cues, to group people across your library, rather than relying on traditional facial recognition. We do not build biometric databases, compare your data against external records, or share identity information between accounts.