Hivemind is DeepLake’s shared memory layer for AI agents. It auto-detects assistants such as Claude Code, OpenClaw, Codex, Cursor, Hermes, and pi, installs hooks, and captures prompts, tool calls, and responses into a shared Deeplake workspace. Teams get BM25 full-text recall, optional semantic retrieval, session summaries, org/workspace isolation, access control, auditability, and BYOC storage across GCS, Azure Blob, AWS S3, or S3-compatible environments.
Fonctionnalités
- Multi-agent auto-detection and hook installation
- Automatic prompt/tool-call/response capture
- Persistent shared team memory
- BM25 and optional semantic retrieval
- Org and workspace isolation
- BYOC cloud storage support
Cas d'usage
- Shared context for AI coding agents
- Team session memory capture
- History recall across Claude/Codex/Cursor
- Reducing repeated token usage
- Agent workspace permission isolation
- Enterprise-owned cloud memory storage
FAQ
Hivemind is DeepLake’s shared memory layer for AI agents. It auto-detects assistants such as Claude Code, OpenClaw, Codex, Cursor, Hermes, and pi, installs hooks, and captures prompts, tool calls, and responses into a shared Deeplake workspace. Teams get BM25 full-text recall, optional semantic retrieval, session summaries, org/workspace isolation, access control, auditability, and BYOC storage across GCS, Azure Blob, AWS S3, or S3-compatible environments. Core capabilities include: Multi-agent auto-detection and hook installation, Automatic prompt/tool-call/response capture, Persistent shared team memory.
Common scenarios include: Shared context for AI coding agents, Team session memory capture, History recall across Claude/Codex/Cursor.
