Memory and context for agents: 9 companies ranked

Companies selling memory layers and context infrastructure that let AI agents remember and retrieve, in SalesBook agentic-sector rank order. Every company has a profile with dated public sources and a live check of its agent interfaces (MCP, llms.txt). Ranks use the published SalesBook method; tell us who is missing.

• Weaviate#13Open-source vector database and AI platform for search, RAG and agents, offered as shared or dedicated cloud to AI teams and enterprises.
  MCP connector: nollms.txt: yes
• Zep#29Context and memory layer for enterprise AI agents built on its open-source Graphiti graph engine.
  MCP connector: yesllms.txt: yes
• Pinecone#49Managed vector database and knowledge platform for developers and enterprises building AI agents, search and recommendation systems.
  MCP connector: nollms.txt: no
• Cognee#57Open-source memory engine for AI agents, with managed Cognee Cloud and an Enterprise tier offering BYOC deployment.
  MCP connector: nollms.txt: no
• Contextual AI#61Context layer platform and RAG APIs for enterprises building AI agents over technical documents, aimed at finance, manufacturing and legal.
  MCP connector: nollms.txt: yes
• Letta#68Platform and open-source framework (from MemGPT) for building stateful AI agents with long-term memory.
  MCP connector: nollms.txt: yes
• Chroma#69Open-source search and retrieval database for AI applications, sold as Chroma Cloud and Enterprise (BYOC) to developers and enterprises.
  MCP connector: nollms.txt: no
• Mem0#77Memory layer for AI agents and apps, offered as open source and a managed platform with free and paid plans.
  MCP connector: yesllms.txt: yes
• Supermemory#109Memory and context API for AI agents, delivered via API, plugins and MCP, with Scale and Enterprise tiers for teams building agents.
  MCP connector: yesllms.txt: yes