Remembra

What is Remembra?

Remembra is a self-hosted persistent memory layer for AI applications that provides entity resolution, temporal decay, graph-aware recall, and hybrid search. It is designed for developers building AI chatbots, agents, and assistants that need to remember facts, decisions, and context across sessions without vendor lock-in.

How to use Remembra?

Install with one command (curl -sSL ... | bash) or via Docker Compose, then use the REST API, Python SDK (pip install remembra), TypeScript SDK (npm install remembra), or MCP server (remembra-mcp) to store and recall memories. Configure the MCP server by adding the remembra MCP tool to Claude, Cursor, or other compatible clients using environment variables like REMEMBRA_URL.

Key features of Remembra

• Smart LLM-powered fact extraction from raw text
• Entity resolution merging aliases into one identity
• Bi-temporal knowledge graph with point-in-time queries
• Hybrid semantic + keyword search for accurate recall
• Built-in PII detection, anomaly monitoring, and audit logging
• 11 MCP tools including store, recall, update, timeline, and share

Use cases of Remembra

• Give a chatbot persistent memory across user sessions without repeating questions
• Enable an AI coding agent to recall past decisions and project context
• Build a personal assistant that remembers preferences, contacts, and schedules
• Create multi-agent systems with shared memory spaces for collaborative reasoning
• Power temporal queries like “Where did Alice work in January 2022?”

FAQ from Remembra

How does Remembra compare to alternatives like Mem0, Zep, or Letta?

Remembra offers free bi-temporal relationships, entity resolution, built-in PII detection, hybrid search, 6 embedding providers, and an MCP server with 11 tools — features often gated behind paid plans or missing in other solutions. See the full feature comparison table in the README.

What dependencies or runtime does Remembra require?

Remembra uses Qdrant for vector storage, SQLite for metadata/graph storage, and optionally Ollama for local embedding inference. It can be run via Docker Compose or the quickstart script; no external API keys are needed for local use.

Where does my data live?

Data is stored locally in your self-hosted instance (Qdrant and SQLite). You control all data; no data is sent to external services unless you configure an external embedding provider.

What transport and authentication does the MCP server support?

The MCP server communicates via REST API over HTTP. Authentication is handled through the REMEMBRA_URL environment variable; local setups typically run without authentication, but enterprise deployments can integrate security features like audit logging and field encryption.

Is Remembra open-source and what license does it use?

Yes, Remembra is released under the MIT License, allowing free use, modification, and distribution.