ByteRover

ByteRover is a local-first, portable memory system for AI agents and developers that replaces vector-based memory with a tiered file-search retrieval pipeline. It organizes context into a hierarchical knowledge tree using stateful memory curation, enabling both agents and humans to reason over and manage memory as a structured knowledge system. ByteRover achieves 92.2% retrieval accuracy on the LoCoMo benchmark and works with any LLM provider via your own API key. Memory runs entirely on your machine by default, with optional cloud sync for portability across machines, teammates, and agent setups.

• Local-first memory — Install via CLI (curl -fsSL https://byterover.dev/install.sh | sh) and run everything on your machine with no account, no cloud, and zero telemetry required.
• Hierarchical knowledge tree — ByteRover curates memories into a structured tree with natural language format, making context queryable and manageable at scale.
• Tiered retrieval pipeline — Replaces vector search with fuzzy text → LLM-driven deep search for higher precision, achieving 92.2% retrieval accuracy on LoCoMo.
• Bring Your Own Key — Power ByteRover with any LLM provider (OpenAI, Anthropic, Gemini, etc.) using your own API key for full model and cost control.
• Cloud sync & portability — Push your local memory workspace to ByteRover Cloud when you need to share across machines, teammates, or agent setups; version-controlled and editable.
• OpenClaw integration — All OpenClaw agents share a persistent, hierarchically structured memory layer through ByteRover.
• Markdown migration — Import existing memory files (MEMORY.md, QMD, any text files) and ByteRover organizes them into a queryable knowledge tree.
• Enterprise-grade security — SOC 2 Type II certified, AES-256 encryption at rest, TLS 1.2+ in transit, RBAC, data residency controls, and on-prem gateway options.
• Team collaboration — Share context across teams like version-controlling code, with cloud sync, context management, and role-based access.