closedloop-technologies / tavily-search
Install for your project team
Run this command in your project directory to install the skill for your entire team:
mkdir -p .claude/skills/tavily-search && curl -o .claude/skills/tavily-search/SKILL.md https://fastmcp.me/Skills/DownloadRaw?id=323Project Skills
This skill will be saved in .claude/skills/tavily-search/ and checked into git. All team members will have access to it automatically.
Important: Please verify the skill by reviewing its instructions before using it.
Use Tavily Search API for optimized, real-time web search results for RAG. Requires TAVILY_API_KEY.
0 views
0 installs
Skill Content
---
name: tavily-search
description: Use Tavily Search API for optimized, real-time web search results for RAG. Requires TAVILY_API_KEY.
---
# Tavily Search Skill
This skill utilizes the Tavily Search API, providing clean, real-time web search results optimized for LLMs and RAG pipelines.
## Setup
1. **Dependencies:** Requires `tavily-python`.
```bash
pip install tavily-python python-dotenv
```
2. **API Key Configuration:** Requires `TAVILY_API_KEY`.
```bash
# If the script fails due to a missing key, run the following:
echo "It seems the Tavily API key is not set up."
read -p "Enter your Tavily API key: " TAVILY_KEY
echo "TAVILY_API_KEY=$TAVILY_KEY" >> .env
if [ -f .gitignore ] && ! grep -q ".env" .gitignore; then echo ".env" >> .gitignore; fi
echo "API key saved to .env."
```
## Usage
Use the `scripts/tavily_search.py` script.
### Command
```bash
python3 scripts/tavily_search.py --query "<query>" [--max-results <N>] [--search-depth <basic|advanced>]
```
### Parameters
* `--query` (Required): The search query.
* `--search-depth` (Optional): Default `basic`. Use `advanced` for intensive research (higher quality, slower).
* `--max-results` (Optional): Default 10.
### Example
```bash
python3 scripts/tavily_search.py --query "autonomous research agents comparison" --search-depth advanced
```
## Output
The script outputs JSON containing a synthesized `answer` (if requested by the script) and a list of `results` (URL, title, content snippets).