A memory layer for AI agents that uses a hierarchical semantic pyramid for long-term context and Mermaid-based symbolic compression for short-term task state. It defaults to a local SQLite backend but requires an external OpenAI-compatible LLM and embedding model for extraction and retrieval operations.
Project overview
Addresses token waste and context loss in long-horizon agent tasks by offloading verbose tool logs to files and structuring recall into layers from conversations to personas.
Project type
AI Agent · Vector Database
Deployment
Refer to project documentation
License
License pending
Best for
Developers and AI engineers building agent systems who need to reduce token usage and preserve context across multi-session workflows.
Key capabilities
Builds a semantic pyramid from L0 Conversations to L3 Personas instead of flat vector piles.
Offloads heavy tool logs to external files and condenses task state into a lightweight Mermaid canvas.
Defaults to a local SQLite + sqlite-vec backend with zero configuration.
Supports both keyword and semantic recall using BM25 + vector + RRF fusion.
Exposes capture, search, and recall HTTP endpoints via a sidecar Gateway service.
Cross-Agent, cross-framework, and cross-device memory import, export, and live migration.
Limitations and risks
Requires OpenClaw version 2026.3.13 or higher.
The Gateway service may issue a security warning if it is exposed without an API key configured.
Operation requires an external OpenAI-compatible LLM and an embedding model for extraction and retrieval tasks.
Getting started
Setup difficulty is rated medium because some configurations require installing dependencies and running patch scripts. The documented first success path involves installing the plugin via OpenClaw, configuring the provider, and verifying the Gateway.
Alternatives and comparisons
An open-source platform that provides an intuitive visual interface for AI workflows, RAG pipelines, and agent capabilities.
Simplifies AI application development by helping developers chain together interoperable components and third-party integrations.