memory-lancedb-pro is an OpenClaw plugin that provides long-term memory capabilities for AI agents using LanceDB as its storage backend.
The plugin solves the problem of AI agents losing context between sessions by automatically capturing and storing agent interactions, then retrieving relevant memories when needed. It uses LLM-powered classification to extract six types of information—profiles, preferences, entities, events, cases, and patterns—and applies a Weibull decay model so important memories persist while noise naturally fades. Retrieval combines vector search with BM25 full-text search, then reranks results using a cross-encoder to surface the most relevant memories before each agent response.
The tool suits teams building OpenClaw agents that need to maintain continuity across conversations and users. It works with any embedding provider including OpenAI, Jina, Gemini, and Ollama, and offers multi-scope isolation so memories can be bounded per-agent, per-user, or per-project. The plugin includes a management CLI for backup, migration, export, and import operations. Developers should note that it requires OpenClaw 2026.3 or later and uses the new before_prompt_build hooks rather than the deprecated before_agent_start approach.
The project maintains active development with recent updates to align with OpenClaw's plugin architecture changes. Documentation is available in multiple languages, indicating sustained internationalization effort. The repository includes a diagnostic command for verifying compatibility after upgrades, suggesting attention to smooth deployment workflows.
