Memori

Memori is a memory infrastructure layer for AI agents that is not tied to any one model provider. It turns agent execution and conversation into structured, persistent state, so production systems can recall what happened across sessions without sending full history to the model each time. Each chat turn is captured and classified into facts, preferences, rules and summaries, and the user controls what is stored, for how long and where. At recall time, only the relevant memories are pulled across conversations and documents, with semantic search added when wording is fuzzy. Each result carries an explanation of why it was included, traceable by entity, time and source. Memori also records the agent's actions, tool calls, decisions and outcomes rather than the conversation alone. It integrates with OpenClaw, Hermes Agent, Claude Code and any agent that supports MCP, and the SDK can be added to existing code with a single line. The vendor reports a score of 87 percent on the LoCoMo benchmark with about 97 percent fewer tokens than full context retrieval. Pricing starts with a free self hosted open source option where you bring your own database, with the full SDK and MCP server. A managed Cloud Free tier costs $0 and includes 5,000 memories created and 15,000 recalled. Production plans are sold on annual contracts: Team from $60K per year for a single production agent on managed multi tenant cloud, and Business from $150K per year for multiple agents on single tenant cloud. Enterprise is custom priced, with customer VPC or on premises deployment.