YouTubeVideoMP4 · MOV · WEBM
Your AI forgets. Soontra remembers.
The memory for everything your team has filmed, photographed and recorded. Search it in Soontra, or from Claude and ChatGPT.
Works with
Search the memory
Files
A memory that grows.
Today it remembers what your team made. Next, it learns when to use it.
Sample workspace- Today
What you made
Every video, photo, recording and document, down to the moment.
- Today
What performed
Post results sync back to the clips behind them.
- Coming next
When to use it
It learns your team’s context, so your AI brings back the right moment.
It remembers every file.
Video, photos, audio and documents, remembered by what’s shown, said and written in them.
PhotosJPG · PNG · HEIC
What’s in the frame
AudioMP3 · WAV · M4A
00:11Every visual moment is captured, so it’s searchable by AI.
Every spoken word
DocumentsPDF · DOCX · TXT
Page by page
One search across all of it
Type a word. Get the moment, the quote and the page.
Give your AI a memory.
Connect Claude, ChatGPT or Cursor once. Your AI can recall any moment and leave drafts. It never publishes.
- Claude
- ChatGPT
- Cursor
- VS Code
Setup for each app
Connect to AI
AI clients connect over MCP, act on your behalf and inherit your role. They can read this workspace, draft with the Write scope, and never delete anything.
MCP server URL
https://app.soontra.com/api/mcpConnects via sign-in (no API key needed).
OAuthClaude Pro/Max or Team/EnterprisePublic HTTPS
Setup steps
- Team/Enterprise: an Owner adds it from Organization settings > Connectors > Add > Custom > Web. Pro/Max: open Customize > Connectors, click "+", then Add custom connector.
- Paste the MCP server URL above as the Remote MCP server URL, then click Add.
- Claude registers automatically when the connector is added.
- Click Connect and sign in to Soontra to authorize this workspace.
- In a chat, open the "+" menu, choose Connectors, and enable Soontra for that conversation.
Test after connecting
List my Soontra projects.Claude, ChatGPT, Cursor and VS Code belong to their owners. Soontra connects over the open Model Context Protocol and is not affiliated with them.
Built for long footage.
Up to 8 hours a video, remembered moment by moment.
- 8 hoursper video, every scene and word
- 5–30 secper moment a search lands on
- 27 formatsof video, photo, audio and document
- Soontra8 hEvery plan
- TwelveLabs4 hMarengo 3.0 search index
- Air4 hVideo Intelligence
- Gemini API3 hPer request, low resolution
- Soontra50 GBEnterprise; 25 GB on other plans
- Air10 GBVideo Intelligence
- TwelveLabs4 GBMarengo 3.0 search index
- Gemini API2 GBFiles API, kept 48 hours
TwelveLabs: Marengo 3.0 · Air: Video Intelligence · Gemini API: video understanding · Gemini API: Files
| Capability | SoontraMedia memory | TwelveLabsVideo AI | Moments LabVideo AI | AirCreative assets | Frame.ioVideo review | GleanWork search | Dropbox DashWork search | Gemini NotebookAI notebook |
|---|---|---|---|---|---|---|---|---|
| What it searches | ||||||||
| Video: what is shown | Yes | Yes | Yes | Yes | PartlyBeta, Team plans up | NoMetadata only | NoMetadata only | NoNo video files |
| Video: what is said | Yes | Yes | Yes | Yes | Yes | PartlyZoom and Gong calls | Not documented | PartlyYouTube captions |
| Images | Yes | Yes | Not documented | Yes | PartlyBeta | PartlyText only, if enabled | Yes | Not documented |
| Audio files | Yes | PartlyDeveloper API | Not documented | Yes | Yes | PartlyCall transcripts | Not documented | Yes |
| PDF and Word documents | Yes | PartlyPDF, developer API | No | PartlyText in PDFs | NoFilename only | Yes | Yes | Yes |
| What it does | ||||||||
| Your whole archive, one search | Yes | Yes | Yes | Yes | PartlyAI search: 30 days | Yes | Yes | NoUp to 600 sources |
| Ties results to the work | YesVia your posts | No | No | PartlyMeta ads, Enterprise | No | No | No | No |
| Your AI can use it over MCP | YesMoments to PDF pages | YesVideo and photos | NoKeeps video from outside AI | PartlyBusiness plans up | PartlyPublic beta | YesCompany text | YesFiles and transcripts | Not documented |
From each product’s public docs, checked October 17, 2026. “Not documented” means we couldn’t confirm it. Soontra transcribes English only today.
Sources
- TwelveLabsMarengo 3.5 modelJockey agentMCP server
- Moments LabSearch and discovery
- AirSummer releaseCreative performanceMCP server
- Frame.ioAI searchMCP server
- GleanSupported file typesZoom transcriptsMCP servers
- Dropbox DashSearch and exploreDash MCP server
- Gemini NotebookSourcesRename to Gemini Notebook
Commonquestions.
What does Soontra do?
It’s a memory for your team’s media. It remembers what’s in every video, photo, recording and document, so you or your AI can find any moment.
Does it edit my footage?
That isn’t the focus. Soontra is the memory: it finds the exact clip, line or page, so you, your editor or your AI can use it.
Can we use it from Claude or ChatGPT?
Yes. Connect once and your AI can search your media, read transcripts and pages, and leave drafts. It never publishes or deletes. In ChatGPT, full read and write needs a Business, Enterprise or Edu workspace. How connections work
How does our archive get in?
Upload it, or import it from Google Drive or Instagram. In a pilot, we bring it in with you.
Is our archive private?
Yes. Your files stay in your workspace, seen only by people you invite. The AI services we use don’t train on customer data.
I’m a solo creator. Can I use Soontra?
Yes. Request beta access and we’ll email you when there’s room. Request beta access
Apply for a pilot.
We’re onboarding a few media teams, with hands-on help from the people building Soontra.
- Apply
- Talk to us
- Bring your archive in
- Put it to work