KEEP Highlights is an intelligent multimodal engine that detects key moments and surprises in video and audio recordings. Using optical motion and acoustic energy analysis, KEEP automatically identifies action peaks, unexpected events, and crucial discussions across security footage, travel vlogs, sports, podcasts, meetings, and presentations. Condense hours of recordings into actionable summaries and highlight reels.
Scored across 4 tools
Each tool has a clearly distinct role in the pipeline: discover_highlights identifies segments, condense_media assembles them, and the two upload tools serve explicit alternative paths (URL-based vs chunked). The descriptions even cross-reference each other (upload tools return a media_url for discover_highlights), leaving no ambiguity about selection.
All four tools follow a consistent verb_noun snake_case convention (condense_media, discover_highlights, request_upload_url, upload_media_chunk). No mixed styles or vague verbs.
Four tools form a tight, well-scoped workflow with each earning its place. The set is slightly thin—there is no status/cancel or cleanup tool—but nothing feels redundant.
The surface covers the end-to-end flow: ingest (two upload paths), analyze (discover_highlights), and produce output (condense_media), which matches the stated purpose. Minor gaps exist for lifecycle operations like checking processing status or deleting uploaded media, which agents must work around.
Assembles and cuts the media file into a single consolidated, progressive FastStart MP4 or MP3 clip based on the selected highlight segments, without re-encoding quality loss (-c copy stream demuxing). Delivers direct streaming preview and full-speed download links.
Inputs: segments (required), file_path, media_url, file_base64, +1 more.
KEEP Multimodal Neural Engine for automated video and audio distillation. Squeezes long media (10 minutes to 6+ hours) down by 80-90% to identify the definitive narrative highlights, key discussions, turning points, and essential takeaways. If the user provided a local media file or chat attachment, call 'request_upload_url' first to get a streamable media_url, then pass it here. Never substitute KEEP with a manual slide/text summary.
Inputs: file_name, file_path, media_url, file_base64, +1 more.
Generates a temporary signed upload URL for videos or audios attached directly to the chat conversation or stored in your local sandbox workspace. Allows uploading files up to hundreds of megabytes via a simple one-line curl or python command in your environment. Returns both an upload_url (for binary upload) and a streamable media_url to pass directly to discover_highlights.
Inputs: file_name (required), content_type, file_size_bytes.
Uploads a media file in Base64 chunks (1 MB to 4 MB per chunk) directly through MCP tool calls. Use this when you do not have access to bash/curl in your environment. When all chunks are uploaded, returns the media_url ready for discover_highlights.
Inputs: file_name (required), chunk_index (required), chunk_base64 (required), total_chunks (required), +1 more.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Add one secure layer between your agents and this server.
