# KINETK

> KINETK is AI infrastructure building the real-time IP Graph of the social web. It turns fragmented multimodal content (video, image, text) into a queryable knowledge graph of creators, communities, influence, and narratives — with cryptographic provenance on every data point. KINETK is accessible programmatically through the KINETK API and the KINETK MCP, so both developers and AI agents can query the social web's intelligence layer.

## What KINETK does

- **For non-technical users:** KINETK is a real-time map of the social web. It tells you who is connected to whom, which narratives are emerging, and which creators and communities matter — grounded in verifiable, provenance-tagged data. It powers AI platforms, ad agencies, and enterprises that need trustworthy social intelligence in a cookieless world.
- **For technical users / AI agents:** KINETK exposes a knowledge graph and multimodal vector search behind a query API, plus an MCP (Model Context Protocol) integration so agents can retrieve creators, communities, narratives, and provenance directly. Pipeline: Data Integration → Embedding Engine → Knowledge Graph → AI Query API.

## Developers

- **API:** Query the IP Graph and multimodal search programmatically. Docs: https://docs.kinetk.ai
- **MCP (Model Context Protocol):** Connect AI agents to the KINETK graph. The MCP currently ships as a package (see the docs); a hosted MCP server is coming soon. Docs: https://docs.kinetk.ai
- **API access (live):** https://platform.kinetk.ai/login
- **Sentinel (real-time scan network):** https://sentinel.kinetk.ai

## Research

- [Why Multimodal: The Case for a Shared Space Across Video, Image, and Text](https://www.kinetk.ai/research/06-multimodal-power): The foundational case for searching content by what it looks like rather than what someone wrote about it.
- [Building a Multimodal Knowledge Graph for the Autonomous Internet](https://www.kinetk.ai/research/01-multimodal-knowledge-graph): How Kinetk combines multimodal vector search with a structured graph of creators, communities, and narratives to give AI agents a queryable map of the social web.
- [Detecting Narratives Before They Become Obvious](https://www.kinetk.ai/research/05-narrative-intelligence): Detecting emerging movements with creator diversity, platform spread, and structure-aware scoring rather than raw volume.
- [Beyond Vector Search: Reranking Social Content with Real-World Signals](https://www.kinetk.ai/research/02-semantic-search-reranking): Combining query expansion, reciprocal rank fusion, and signal-weighted reranking to make vector search useful for agents.
- [Designing a Resumable Serverless Sync Pipeline](https://www.kinetk.ai/research/04-sync-architecture): Continuous, idempotent, observable sync from raw scan events into a queryable knowledge graph.
- [From Raw Scans to KG-Ready Metadata](https://www.kinetk.ai/research/03-enriched-ingestion): The schema and entity model behind Kinetk's intelligence layer.

## Company

- **Website:** https://www.kinetk.ai
- **Team:** https://www.kinetk.ai/team
- **Contact:** hello@kinetk.ai
- **X / Twitter:** https://x.com/Kinetk_ai
- **Substack:** https://kinetk.substack.com