Berlin startup Cognee secured €7.5 mn to expand its structured memory infrastructure for AI systems and autonomous agents.
Pebblebed led the round, with 42CAP returning as a follow-on investor alongside several new and existing angels.
Founded in 2024, the Berlin-based company develops infrastructure addressing a persistent weakness in modern AI architectures: lack of durable, structured memory.
Most deployed systems operate in stateless modes, relying on file retrieval pipelines or narrow context windows. Those approaches strain under complex, ongoing enterprise workflows.
Cognee converts unstructured data into persistent memory built on knowledge graphs and semantic models. The platform stores contextual relationships rather than isolated text fragments.
AI agents built on top of this layer retain historical interactions, reason across linked datasets, and reduce hallucination risk in production settings.
According to Beinsure analysts, memory persistence increasingly defines enterprise-grade AI reliability.
The company launched as an open-source project and gained traction among developers building AI-native applications.
Thousands of engineers utilize the tooling to embed contextual awareness into internal copilots, workflow automations, and domain-specific agents. Open distribution accelerated adoption before formal commercialization.
Although headquartered in Berlin, Cognee operates across Europe and the US. Management positions the company as foundational infrastructure rather than an application vconcludeor.
CEO Vasilije Markovic argues model capability no longer acts as the primary constraint. Systems fail when they cannot maintain continuity across tinquires and time horizons.
The new capital will fund product development, scaling of enterprise support, and expansion of its global developer ecosystem.
We believe structured memory layers will define the next phase of AI infrastructure as enterprises push agents into longer-running, business-critical processes where stateless retrieval falls short.
Leave a Reply