How to Install and Use Tavily AI Search in OpenClaw

ClawHub is OpenClaw's official skill registry — the equivalent of npm for OpenClaw capabilities. Installing the tavily-ai-search skill adds web search functionality to your OpenClaw instance in a single command. Open your terminal (or the terminal panel inside your OpenClaw interface) and run the install command below. The installation downloads the skill package, validates its integrity, and registers it with your local OpenClaw configuration. You will see a confirmation message showing the skill version installed and the configuration keys it requires. The skill will not activate until you supply the API key in the next step — attempting to use it before configuration will produce a clear error message pointing you to the missing key. If you see a `clawhub: command not found` error, your OpenClaw installation may be incomplete or the ClawHub binary may not be on your PATH. Check your OpenClaw installation documentation to verify the installation path, or reinstall OpenClaw to get a fresh ClawHub setup.

1clawhub install tavily-ai-search

Tavily provides a free API tier with 1,000 requests per month — plenty for personal use and development. Navigate to app.tavily.com and sign up or log in. Once inside the dashboard, look for the 'API Keys' section in the left sidebar or under your account settings. Click 'Create API Key'. Give it a descriptive name like 'OpenClaw Integration'. Tavily generates the key immediately — it starts with `tvly-`. Copy the full key now, as the dashboard may not display it again after you navigate away (though you can always generate a new one if needed). Tavily API keys on the free tier are rate-limited to 100 requests per minute and 1,000 per month. If you plan to use Tavily AI Search heavily or in automated workflows, consider upgrading to a paid plan. The Research plan at $35/month provides 10,000 requests per month with higher rate limits and access to advanced search depth settings. Keep your API key private — it is scoped to your Tavily account and any searches performed with it count against your quota.

OpenClaw stores skill configuration in a local config file or as environment variables, depending on your installation type. The recommended method is to set the key in your OpenClaw `.env` file or skill config — this keeps credentials out of shell history and applies them persistently across sessions. Locate your OpenClaw configuration directory (typically `~/.openclaw/` on macOS/Linux or `%APPDATA%\openclaw\` on Windows). Open the `.env` file in that directory with any text editor, or create it if it does not exist. Add your Tavily API key as shown in the code block below. Alternatively, use the `clawhub config set` command to set the key directly from the terminal — this writes to the same config file without requiring you to open an editor. After saving the configuration, restart OpenClaw or reload your configuration for the change to take effect. On most installations, running `clawhub reload` is sufficient without a full restart.

1# Option A: Add to ~/.openclaw/.env
2TAVILY_API_KEY=tvly-your-api-key-here
3
4# Option B: Set via clawhub config command
5clawhub config set TAVILY_API_KEY tvly-your-api-key-here
6
7# Verify the key is set
8clawhub config get TAVILY_API_KEY

Open OpenClaw chat and type a natural language search prompt. The tavily-ai-search skill activates automatically when OpenClaw determines a web search is appropriate — you do not need to prefix commands or invoke the skill by name. Just ask questions naturally and OpenClaw routes search-type queries to Tavily. Tavily AI Search returns results with the following structure: a synthesized answer (when available), a ranked list of sources with titles, URLs, and extracted content snippets, and relevance scores for each result. The AI post-processing phase is what produces this structured output — raw Tavily Web Search returns the same sources but without the ranking and synthesis layer. For optimal results, be specific in your queries. Vague queries like 'tell me about AI' produce lower-quality results than targeted queries like 'what are the latest benchmark results for Claude 3.7 Sonnet on coding tasks'. The more context Tavily has, the better it can rank and filter results.

The tavily-ai-search skill exposes several configuration options that let you tune search behavior for your specific use cases. The two most important settings are search depth and max results. **Search depth** controls how thorough Tavily's crawling is. The default `basic` depth fetches the top search results and extracts content from their main pages. The `advanced` depth performs a deeper crawl, following links and extracting content from more pages per result — useful for research tasks but slower and more expensive in API credits. **Max results** controls how many ranked sources are returned per query. The default is 5, which is appropriate for most conversational queries. For research tasks where you want broad coverage, increase this to 10. For quick factual lookups, reduce it to 3 to get faster responses. You can also set a time-based filter to restrict results to content published within a specific timeframe — useful for news monitoring and keeping OpenClaw's answers current. RapidDev offers custom OpenClaw configuration reviews for teams looking to optimize skill settings for specific research workflows — particularly useful when integrating Tavily AI Search into automated pipelines alongside other skills.

1# Advanced skill configuration in ~/.openclaw/skills/tavily-ai-search.yaml
2tavily-ai-search:
3  search_depth: advanced      # basic (default) or advanced
4  max_results: 10             # default: 5, range: 1-20
5  include_answer: true        # include synthesized answer (default: true)
6  include_raw_content: false  # include full page content (increases response size)
7  days: 7                     # restrict to content from last N days (optional)
8  include_domains: []         # restrict to these domains only (optional)
9  exclude_domains: []         # exclude these domains (optional)