Search inside your video, images, audio and documents
Most of a company's data sits in object storage as video, images, audio and documents, and almost none of it is searchable by what is inside.
Mixpeek connects to your bucket and runs extractors over each file. A video becomes scenes, keyframes, transcripts, faces and on-screen text. A PDF becomes pages, tables and figures. Each piece is embedded and stored with its timestamps and source path.
Then you query it. Retrievers combine vector search, filters and reranking in one call, so "red car at night near the stadium" returns the exact second of footage. Taxonomies label files against your own categories. Clusters group what nobody has labeled yet.
It ships as an API, Python and JavaScript SDKs, an MCP server for agents, and Studio, a web app where you can index a folder and search it in a few minutes.
Pricing starts at $25 a month plus usage.