An intelligent web search skill that uses local Retrieval-Augmented Generation to rank and prioritize search results from multiple engines.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install local-rag-search
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 local-rag-search 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 Local RAG Search?
Local RAG Search is a sophisticated skill designed for Openclaw Skills that integrates the mcp-local-rag server to perform intelligent web queries. Unlike standard search tools, it utilizes semantic similarity ranking to ensure the most relevant information is prioritized, all without relying on proprietary search APIs for the ranking logic. It provides a bridge between live web data and local AI processing, ensuring that the context provided to your model is both current and highly relevant.
This skill is essential for developers and researchers who need high-quality context from the internet while maintaining control over how data is processed and ranked. By leveraging engines like DuckDuckGo, Google, and Wikipedia, it offers a versatile toolkit for everything from quick fact-checking to deep, multi-perspective investigations. It effectively turns a standard search into a local RAG pipeline for your agent.
Local RAG Search Use Cases
- Researching complex technical topics that require diverse perspectives from multiple search backends.
- Performing privacy-conscious web searches using DuckDuckGo for general information gathering.
- Gathering comprehensive context for LLMs to reduce hallucinations by providing semantically relevant snippets.
- Conducting deep-dive investigations into specific events or scientific documentation using multi-engine search capabilities.
How Local RAG Search Works
- The user provides a natural language query or a set of search terms to the agent utilizing Openclaw Skills.
- The skill invokes the mcp-local-rag server, which fetches raw results from selected backends like Google, Brave, or Wikipedia.
- Local semantic similarity models rank the fetched results based on their relevance to the original query context.
- The skill filters the results to return the top-k most relevant snippets, including source URLs and metadata.
- The AI agent synthesizes the ranked information to provide a grounded, evidence-based response to the user.
Local RAG Search Setup
To use this skill within the Openclaw Skills ecosystem, you must have the mcp-local-rag server configured in your environment.
# Ensure the MCP server is available in your configuration
# The skill will automatically interface with the following tools:
# - rag_search_ddgs
# - rag_search_google
# - deep_research
Configure your agent to point to the local-rag-search skill definition and ensure your network allows the MCP server to reach the designated search backends.
Local RAG Search Data Schema & Taxonomy
The skill organizes search data into structured objects that include semantic relevance scores and source tracking.
| Property | Description |
|---|---|
| query | The natural language string or search terms used for the lookup. |
| backend | The specific engine used (e.g., google, duckduckgo, wikipedia). |
| num_results | The total number of raw results fetched before ranking. |
| top_k | The number of semantically relevant results returned to the agent. |
| include_urls | Boolean flag to include or exclude source links in the output. |
Local RAG Search Advanced Features
- Multi-engine deep research allowing simultaneous queries across Google, Bing, and Wikipedia for broad coverage.
- Semantic similarity ranking performed locally to ensure data relevance without external ranking bias.
- Customizable backends including niche engines like Mojeek and Grokipedia for varied perspectives.
- Privacy-first defaults within Openclaw Skills that prioritize DuckDuckGo to minimize user tracking.
- Batch processing support for multiple related search terms in a single deep research lifecycle.
SKILL.md
Loading
Related Openclaw Skills
Prediction Market Strategy Bot (Dawn CLI)
Execute the complete Dawn CLI strategy lifecycle from authentication and creation to live deployment and monitoring.
Grok Imagine Image Pro
A high-performance image generation and editing tool leveraging xAI Grok and Flux APIs.
Todolist MD Clawdbot
A specialized AI skill for managing Markdown-based todo lists across local storage, Google Drive, and S3 while maintaining line stability and minimal token usage.
Todolist-MD Clawdbot
An AI-powered agent skill that manages Markdown todo files using line-stable edits and specific bot markers for seamless task synchronization.
SURGE OpenClaw (Dev)
SURGE OpenClaw enables AI agents to manage the full lifecycle of tokens, from secure wallet creation and free funding to automated trading on EVM and Solana chains.
Recraft
A professional-grade image generation and editing tool for Openclaw Skills that leverages the Recraft API for raster and vector assets.
Featured*
