Video Searching Agent
Open-source + managed API. Multi-platform discovery across YouTube, TikTok, Instagram, and X — query parsing, ranked references, SSE streaming.
Video Editing Agent (VEA)
Open-source + managed primitives. End-to-end pipeline: indexing → script generation → clip planning → TTS → render. Or call
/video/edit, /video/clip, /video/split directly.Screenplay Extraction
Managed API only. Submit a short drama episode → get back structured storyboard data (scene breakdowns, transcripts, character namelist). Async with webhook callback.
Agent Examples
Eight runnable ReAct agents — SOP compliance, service quality, security, public-platform search, video editing, personal memory, visual RAG, creator intelligence. Live-verified against
api.memories.ai.What’s Inside
Looking for the Memories CLI? It’s a separate ecosystem-wide tool — see the Documentation overview.
Which One Do I Use?
Two Ways to Adopt
The managed APIs and the open-source agents are kept in sync — the open-source code is the reference implementation of what the managed endpoint runs.
Shared Foundation
- Underlying APIs: All agents/services call Visual Intelligence (for inference) and optionally Visual Search (for indexed search).
- Auth: Same
Authorization: sk-mavi-...API key as the rest of the platform — no separate key. - Billing: Calls hit the underlying Visual Intelligence / Visual Search endpoints and bill against those line items. No additional “agent fee”.
- Webhooks: Async services (Screenplay Extraction, VEA) require a configured webhook — see Webhooks.
Getting Started
- Pick the agent or service that matches your goal (table above).
- Get an API key from the Memories.ai Console. The same key works across Visual Intelligence, Visual Search, and Visual Agents.
- Configure webhooks if you’ll use any async endpoint (VEA, Screenplay Extraction).
- Open the per-agent page for setup instructions specific to that agent.