The Discovery Agent is your AI-powered research assistant for video. Describe what you're looking for in plain language and it returns matching clips, quotes or moments in seconds.
What it does
The Discovery Agent understands natural language and context, so it can surface moments even where metadata is thin or missing. Our AI indexes every frame and every word in your catalog, and the Agent works from intent, synonyms and related concepts rather than exact matches. Think of it as the colleague who has watched every episode and is always free.
Key benefits
Find hidden gems: unearth forgotten or buried moments in your catalog.
Work faster: skip scrubbing and searching, go straight to the story.
Empower everyone: freelancers, social teams and non-technical users can search with confidence.
Free up your experts: archivists and system managers spend less time fielding requests.
Stay secure: you only see what you're allowed to, and no content leaves your system.
What it can’t do
Find specific spoken words.
The Agent works from meaning, not exact wording. To find every instance of a particular word or phrase, use the search bar with a transcript filter, narrowing by person label if you need to.
Manage your metadata.
The Agent is built for finding content, not maintaining it. It can't fix misspellings, correct transcripts, flag people missing from the thesaurus or count label usage.
Access the Discovery Agent
Click the Discovery Agent icon in the top-right corner of the platform. It's available from anywhere, and opens as a chat panel over your current view, so you don't lose your place. Expand it for a larger view, or close it anytime.
2. Type your request in plain language, the way you'd brief a colleague. Your prompt can be short and simple, or a detailed brief. Learn how to write a great prompt.
Review and use your results
Once your prompt is processed, the Agent returns results that can include precise clips, photos, full videos, or a combination, each paired with an MXT3 description so you know what's in it before you open it. You can interact with results in several ways:
Open in Studio: send a result straight into Studio to start editing for social.
Add to a project: add one or all returned clips to a project for later use.
Download directly: download clips straight to your device.
Find similar content: ask the Agent to find more content similar to a specific result.
View and clip: hover over a result to preview each frame, then open it to see it in full with its metadata, and use the clipping tools to trim to the timecodes you need.
Media context and follow-ups
Open the Agent from a media page and it automatically uses that media as context, tailoring its results without any extra input from you. Remove the context anytime if you want to search more broadly.
The Agent also remembers your previous prompts within a conversation, so you can refine results, like narrowing to a specific time period or content type, without starting over.
Give feedback
Use the thumbs up or thumbs down on any response to tell the Agent whether the results matched what you needed. If something's off, add a quick note on what was missing or incorrect. Over time, this feedback helps the Agent deliver more relevant results tailored to your workflows.
What's next
Need more help? Reach out to our Support team, we're happy to help.
UI not matching? Your platform may not be updated yet, let us know and we'll help.