Memori

Memori is a persistent memory layer for developers building AI agents that need context across sessions. It records conversations and agent activity, including tool calls, decisions and outcomes, so agents can recall prior work and user preferences. Storage options include your own database and the hosted Memori Cloud service.

It organizes interaction history into facts, preferences, rules and summaries, then retrieves relevant context across conversations and documents. Semantic retrieval helps match requests whose wording differs from stored memories. Recall results provide context about relevance, entities, time and source.

Python and TypeScript SDKs connect to applications using OpenAI, Anthropic, Bedrock, DeepSeek, Gemini and Grok. Integrations include Agno, LangChain and Pydantic AI. OpenClaw and Hermes Agent integrations capture memory between sessions; MCP connects coding assistants to persistent project context.

Using your own database controls where memory is stored, but does not by itself establish an offline processing path. Memori's Advanced Augmentation is a separate service that enriches memories in the background, and the README documents service access both with and without an account. Memori Cloud offers hosted storage and search. The project also describes enterprise deployment in a private VPC.

Its website documents database-based access controls, identity provider integration, retention policies and audit records for memory access and deletion.