Cipher

Memory-powered agent framework that provides persistent memory capabilities across conversations and sessions using vector databases and embeddings, enabling context retention, reasoning pattern recognition, and shared workspace memory for team collaboration.

Memory-powered AI agent framework built by BYTEROVER that provides persistent memory capabilities across conversations and sessions using vector databases and embeddings. The implementation supports multiple LLM providers (OpenAI, Anthropic, Gemini, Ollama, etc.), vector stores (Qdrant, Milvus, in-memory), and includes specialized workspace memory for team collaboration tracking project progress, bug reports, and member activities. Features include automatic memory extraction and storage, reasoning pattern recognition, knowledge graph integration, and MCP aggregator mode that exposes memory tools alongside connected MCP servers with conflict resolution, serving developers who need AI assistants that learn and remember context across coding sessions, teams requiring shared project memory, and organizations building AI workflows that benefit from persistent knowledge retention.