A connector-agnostic memory layer utilizing Postgres and pgvector for incremental data ingestion and semantic search capabilities.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install openclaw-universal-memory
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install openclaw-universal-memory using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is OpenClaw Universal Memory?
OpenClaw Universal Memory is a robust, production-ready memory layer designed to bridge the gap between fragmented data sources and AI-driven applications. By leveraging Postgres and the pgvector extension, this skill enables developers to ingest, normalize, and store data from multiple systems into a unified canonical schema. It serves as a foundational component for those building advanced AI workflows using Openclaw Skills, providing a reliable backbone for Retrieval-Augmented Generation (RAG).
The skill excels at handling incremental synchronization through a cursor-based architecture, ensuring that only new or updated records are processed from sources like Gmail, Slack, or custom JSON feeds. This makes it an efficient choice for maintaining long-term agent memory without redundant processing or high token costs. Whether you are aggregating organizational knowledge or personal archives, OpenClaw Universal Memory provides the searchable, high-performance storage needed for context-aware AI agents.
OpenClaw Universal Memory Use Cases
- Normalize and consolidate records from disparate systems like Gmail, Slack, and Asana into a single, searchable schema.
- Maintain efficient incremental synchronization history using cursors to prevent data duplication and reduce API load.
- Create RAG-ready vector storage for AI agents to retrieve contextually relevant information via high-speed semantic search.
- Centralize private organizational knowledge into a self-hosted Postgres database for maximum data sovereignty and privacy.
How OpenClaw Universal Memory Works
- Initialize the Postgres schema with pgvector support to prepare the environment for vectorized data storage.
- Configure source connectors or provide JSON/NDJSON inputs containing the data to be remembered by the system.
- The skill ingests data, mapping it to a canonical format including titles, body text, and comprehensive metadata.
- Embeddings are generated and stored alongside the raw content for high-performance semantic indexing within Postgres.
- AI agents query the memory layer using semantic search to retrieve the most relevant context for specific user prompts.
- Sync cursors are updated per source account to facilitate seamless, incremental updates during the next ingestion cycle.
OpenClaw Universal Memory Setup
Ensure you have a Postgres instance with the vector extension enabled. Then, follow these steps to integrate with your Openclaw Skills workflow:
# Install the package and database dependencies
pip install -e .
pip install "psycopg[binary]>=3.2"
# Configure your Database DSN
python skills/openclaw-universal-memory/scripts/run_memory.py --action configure-dsn
# Initialize the database schema
python skills/openclaw-universal-memory/scripts/run_memory.py --action init-schema --dsn-env DATABASE_DSN
OpenClaw Universal Memory Data Schema & Taxonomy
The memory layer utilizes a standardized contract for all ingested entities to ensure compatibility across different Openclaw Skills:
| Field | Description |
|---|---|
external_id |
Unique identifier from the source system (e.g., Message ID). |
entity_type |
The category of data, such as email, message, or task. |
title |
A short, descriptive title for the record. |
body_text |
The primary text content used for vectorization and search. |
raw_json |
The full original source data for deep inspection. |
meta_json |
Key-value pairs used for filtering and classification. |
next_cursor |
The marker used for incremental synchronization. |
OpenClaw Universal Memory Advanced Features
- Connector-Agnostic Design: Easily extend the system by writing custom adapters for any data source using provided Python templates.
- Incremental Syncing: Native support for cursor-based ingestion, significantly reducing processing time and resource consumption.
- Built-in Diagnostics: Includes 'doctor' checks and connector validation actions to verify authentication and database health.
- Semantic Search: High-performance vector search capabilities that allow agents to find information based on meaning rather than just keywords.
- Automation Friendly: Designed for easy integration with cron jobs and CI/CD pipelines for periodic, hands-off memory updates.
SKILL.md
Loading
Related Openclaw Skills
uv-priority
A mandatory skill for AI agents that replaces traditional pip and python commands with the high-performance uv package manager.
Clean Pytest
A framework for writing maintainable Python tests using Fake-based implementations and contract testing.
Flight Search
A comprehensive flight aggregation and price monitoring tool designed for AI agents to automate travel research.
Bitrefill Website
A navigation and discovery skill for exploring Bitrefill's global catalog of digital goods including gift cards, mobile top-ups, and travel eSIMs.
OpenClaw Web Automation (Basic)
A high-speed, credential-free browser automation tool for scraping, summarizing, and auditing public websites.
clawring
A managed voice service for AI agents to initiate real-world phone calls and two-way conversations.
Featured*
