Founder Adam Ghetti unveils Adapter, a knowledge‑graph infrastructure that lets developers build AI tools with better data control and lower costs.

Adapter has unveiled a new “Cognition” layer designed to give developers tighter control over data while cutting the cost of building AI‑powered applications.

What the Cognition Layer Does

The Cognition layer sits on top of Adapter’s existing knowledge‑graph infrastructure, providing a unified API that abstracts away the complexities of data ingestion, normalization, and retrieval for large language models.

According to founder Adam Ghetti, the layer lets developers specify granular permissions for each data source, ensuring that sensitive information is only exposed to authorized AI agents.

Key Benefits for AI Tool Builders

  • Reduced latency through pre‑computed graph traversals
  • Lower operational costs by eliminating redundant data pipelines
  • Built‑in versioning and audit trails for compliance
  • Plug‑and‑play compatibility with major LLM providers

These features aim to address common pain points such as data silos, unpredictable pricing models, and the difficulty of maintaining up‑to‑date knowledge bases for AI systems.

How It Works

Developers upload raw datasets to Adapter’s platform, where the Cognition engine automatically maps entities, relationships, and attributes into a graph schema. The resulting graph can then be queried in real time via a RESTful endpoint or integrated directly into prompt engineering workflows.

The layer also supports Upstarts Media’s deep‑dive coverage of Adapter, which details the technical architecture and early adopter case studies.

“Cognition gives us the confidence to scale AI features without sacrificing data governance,” said a beta user from a fintech startup.

Adapter plans to roll out additional modules, including automated ontology enrichment and cross‑domain federation, later this year.

Upstarts Media coverage of Adapter’s Cognition launch