llm-tool-exa

Exa.ai integration for the LLM command-line tool.

This plugin gives LLMs the ability to search the web, crawl pages, and find similar content using the Exa API.

Installation

Install this plugin using llm:

llm install llm-tool-exa

To verify the installation, run:

llm plugins
# You should see 'llm-tool-exa' or 'exa.ai api tools' in the list

For development (local install):

llm install -e .

Configuration

You need an Exa API key. Get one here.

Configure it using llm keys:

llm keys set exa
# Enter your key when prompted

Or set the EXA_API_KEY environment variable.

Available Tools

This plugin registers the Exa toolbox. By default, the following tools are enabled:

• search_web: Search the web using Exa's neural or keyword search. Supports filtering by domain, date, category (news, company, paper, etc.), and more.
• get_contents: Retrieve the full text content of specific URLs (web scraping).

Optional Tools

The code supports these additional tools, which may be enabled depending on configuration:

• find_similar: Find websites similar to a given URL.
• answer: Get a direct AI answer with citations.
• research: Conduct a "Deep Research" task (long-running background research).

Usage

Once installed, you can use these tools with models that support function calling (e.g., GPT-5, Claude 3.5 Sonnet).

llm chat -T Exa -m gpt-5 -s "You have access to Exa web search tools.
> Find the latest release notes for the 'uv' python package and summarize the key features."

The model will use search_web to find the page and get_contents to read the details.

Development

This project recommends uv for dependency management.

1. Setup Environment:

   # creates .venv if needed, install dependencies there, and the package in editable mode
   uv sync

2. Run locally in uv-native style within the managed environment:

   uv run llm --help
   uv run llm plugins
   uv run pytest           # run tests within the .env

3. OR manually activate venv shell instead of uv run

   source .venv/bin/activate

   Then use python, pip, llm etc. directly.