A specialized semantic search, clean web scraping, and content extraction skill engineered specifically for AI agents and LLMs.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install exa-search-ws
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 exa-search-ws 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 Exa Search?
Exa Search is a high-performance, custom-built semantic search engine designed to elevate the capabilities of AI agents and LLMs within the Openclaw Skills ecosystem. Unlike traditional keyword-based engines, it leverages neural embeddings to understand the conceptual meaning of complex queries, enabling agents to retrieve highly relevant, high-quality information from across the web.
By integrating this tool, developers can provide their LLMs with boilerplate-free Markdown text, extractive highlights, and similarity-based recommendations. This minimizes token consumption and dramatically improves the accuracy of retrieval-augmented generation (RAG) pipelines, making it an essential building block for constructing advanced search workflows using Openclaw Skills.
Exa Search Use Cases
- Conducting deep, multi-step web research on complex subjects using natural language rather than keyword matching.
- Extracting clean, ad-free Markdown content from articles, documentation, or news archives directly into LLM context windows.
- Implementing Retrieval-Augmented Generation (RAG) patterns where highly relevant extractive highlights are required to save API tokens.
- Building structured datasets or concept lists using conceptual similarity queries based on a starting URL.
How Exa Search Works
- Formulate Query: The AI agent or user submits a semantic search query, a URL to crawl, or a similarity search request.
- Neural Index Look-up: Exa executes a semantic embedding search or similar URL lookup to identify conceptually matched resources.
- Cache or Live Crawl: The skill checks cache freshness configuration (
max_age_hours) to either pull optimized cached pages or perform a real-time live crawl. - Clean & Transform: Extracted pages are stripped of noise (ads, menus, sidebars) and transformed into optimized, clean Markdown or extractive highlights.
- Inject into Context: The agent formats these results (often in structured XML blocks) to use directly in downstream generation or decision-making.
Exa Search Setup
Prerequisites and Installation
Install the Exa python library:
pip install exa_py
Configuration
Set up your environment variable with your API key:
export EXA_API_KEY="YOUR_EXA_API_KEY"
CLI Usage
To run the command-line search utility for testing and manual executions:
/home/ubuntu/skills/exa-search/scripts/exa_search.py --help
Exa Search Data Schema & Taxonomy
Parameter Configuration
| Parameter | Type | Description |
|---|---|---|
query |
String | Natural language search prompt used for semantic mapping. |
urls |
Array of Strings | Specific targets for extraction. |
max_age_hours |
Integer | Controls cache freshness (0 = live crawl, -1 = cache only, >0 = threshold). |
subpages |
Integer | Maximum subpages to crawl on a target site. |
Return Format
- Search Results Schema: Contains a list of result objects, containing
url,title,score, and optionallyhighlights. - Extraction Content Schema: Returns the parameter
textas a cleaned Markdown string andhighlightscontaining query-relevant snippets.
Exa Search Advanced Features
- Subpage Crawling: Discover and index deep hierarchies of nested pages automatically (e.g., target specific API docs or reference directories) with custom subpage limits.
- Cache Freshness Control: Fine-grained optimization using the
max_age_hoursparameter to balance search latency and live correctness. - RAG XML Integration Pattern: Structured template logic to cleanly wrap retrieved contents in standard XML tags, preventing context pollution in downstream generation tasks.
- Similarity Mapping: Retrieve conceptually related pages from a starting URL without having to formulate manual keyword search terms.
SKILL.md
Loading
Related Openclaw Skills
Firecrawl
An AI-native web scraping, crawling, and structured extraction skill that converts websites into LLM-ready data.
Relationship Chat Analysis
An AI agent skill designed to parse, normalize, and analyze long relationship chat exports to extract communication patterns and emotional dynamics with strict privacy safeguards.
Enterprise Law Risk Checker
An expert AI legal compliance assistant engineered to detect hidden corporate regulatory risks and provide actionable rectification plans based on 27 years of senior partner experience.
Fable Explainer
An advanced conceptual teaching engine that leverages original fables, daily analogies, and progressive disclosure to make complex technical topics permanently memorable.
Pronoun Resolver
A zero-latency local context hook that detects ambiguous pronouns, vague referents, and bare imperatives, helping AI agents resolve them accurately based on conversation history.
Tavily Search & Research
An AI-optimized search and extraction skill that connects agents to the web using Tavily's specialized RAG-friendly APIs.
Featured*
