Tavily

OpenCode plugin for Tavily — gives your AI agent reliable web search, content extraction, crawling, URL discovery, and deep research with citations.

Supports OpenCode v1.18.29+ (1.x) and v2 using OpenCode's documented dual-version entrypoint: v1 calls server(), and v2 calls setup().

Installation

Add the plugin to your opencode.json. The configuration key differs between versions.

OpenCode v2

{
  "$schema": "https://opencode.ai/config.json",
  "plugins": ["opencode-tavily"]
}

OpenCode v1.18.29+

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-tavily"]
}

OpenCode installs the package automatically on next launch

Install from Git

Use the Git source instead of the npm package name. For v2:

{
  "$schema": "https://opencode.ai/config.json",
  "plugins": ["github:tavily-ai/opencode-tavily"]
}

For v1.18.29+:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["github:tavily-ai/opencode-tavily"]
}

Install the Tavily CLI

Then install the Tavily CLI globally:

curl -fsSL https://cli.tavily.com/install.sh | bash

Or via Python:

uv tool install tavily-cli
# or
pip install tavily-cli

Authentication

On first use, the agent will prompt you to authenticate. You can also set up in advance:

# Browser login (recommended)
tvly login
# Or set an API key
export TAVILY_API_KEY=tvly-your-api-key

Get an API key at tavily.com.

If TAVILY_API_KEY is set in your environment, the plugin automatically passes it to shell commands.

What It Does

This plugin registers the Tavily CLI skill with OpenCode. Once installed, the agent can:

• Search the web with optional content extraction
• Extract any webpage to clean markdown or text
• Map all URLs on a website
• Crawl entire websites recursively
• Research — AI-powered deep research with citations

All output is written to a .tavily/ directory to avoid flooding context.

Tools Summary

Command	Use When	Key Flags
tvly search	No specific URL yet — find sources and answer questions	--client-name opencode, --depth, --max-results, --time-range, --include-raw-content
tvly extract	Have URL(s) — pull clean content (up to 20 per call)	--client-name opencode, --extract-depth, --query, --chunks-per-source
tvly map	Need to discover URLs on a large site	--client-name opencode, --limit, --instructions, --max-depth
tvly crawl	Need bulk content from a site section	--client-name opencode, --max-depth, --max-breadth, --output-dir
tvly research	Need comprehensive, multi-source analysis	--client-name opencode, --model, --stream, status / poll

Links

• Tavily
• Tavily Documentation
• Tavily CLI Installation
• OpenCode Plugin Docs
• OpenCode Ecosystem

License

MIT