Giving Your AI Real-Time Web Access with OpenClaw

I often find myself frustrated when an AI agent is stuck behind its training cutoff. It knows plenty about the past, but it can’t tell me what happened in a GitHub repo an hour ago or check the latest documentation for a library I’m using. Building a reliable bridge between an LLM and the live web is a common headache for developers.

OpenClaw handles this with two lightweight tools: web_search and web_fetch. They don’t use heavy browser automation, so they stay fast and efficient for most tasks. Here is how I set them up.

What You’ll Need

• An OpenClaw installation.
• An API key from a supported provider: Brave, Perplexity, Gemini, Grok, or Kimi.
• A Firecrawl API key (optional, for better content extraction).

Quick Start

You can get web search running in about five minutes. I usually prefer Perplexity or Brave because they provide structured results quickly.

1. Get your API key: For example, generate one at brave.com/search/api or perplexity.ai/settings/api.
2. Configure OpenClaw: Run the configuration command in your terminal:

   openclaw configure --section web

3. Set the Environment Variable: If you prefer not to use the config file, you can export your key directly:

   export BRAVE_API_KEY="your_key_here"
   # or
   export PERPLEXITY_API_KEY="your_key_here"

4. Test the Tool: Once configured, the web_search tool becomes available to your agent automatically.

Choosing a Search Provider

OpenClaw supports several providers. If you don’t pick one, it tries to auto-detect based on your environment variables in this order: Brave, Gemini, Kimi, Perplexity, then Grok.

Provider	Pros	API Key
Perplexity	Fast, structured; includes filters and extraction	PERPLEXITY_API_KEY
Brave	Fast, structured results	BRAVE_API_KEY
Gemini	Google Search grounding with citations	GEMINI_API_KEY
Grok	xAI web-grounded responses	XAI_XPI_KEY

Configuration Examples

If you want to be explicit, you can define your provider in your configuration file. Here is how I set up Perplexity:

{
  tools: {
    web: {
      search: {
        enabled: true,
        provider: "perplexity",
        perplexity: {
          apiKey: "pplx-...", // optional if PERPLEXITY_API_KEY is set
        },
      },
    },
  },
}

If you prefer Gemini for Google Search grounding, use this structure:

{
  tools: {
    web: {
      search: {
        provider: "gemini",
        gemini: {
          apiKey: "AIza...",
          model: "gemini-2.5-flash",
        },
      },
    },
  },
}

Using the Search Tool

The web_search tool supports several parameters to help you narrow down results. You can filter by date, country, or even specific domains if you use Perplexity.

// Recent results (past week)
await web_search({
  query: "TMBG interview",
  freshness: "week",
});

// Domain filtering (Perplexity only)
await web_search({
  query: "climate research",
  domain_filter: ["nature.com", "science.org", ".edu"],
});

Fetching Page Content

While web_search finds links, web_fetch reads them. It uses a plain HTTP GET request and converts HTML into clean markdown or text. It is enabled by default.

I use it like this:

await web_fetch({
  url: "https://example.com",
  extractMode: "markdown"
});

Keep in mind that web_fetch does not execute JavaScript. If you hit a site that is a heavy Single Page App (SPA) or requires a login, this tool won’t work.

Troubleshooting

Missing API Key If you try to search without a key, OpenClaw returns a short setup hint. You can fix this by running openclaw configure --section web or setting the correct environment variable for your provider.

Empty Content or JS Errors If web_fetch returns empty results or “Readability failed” on a site, it is likely because the site requires JavaScript to render. For these cases, you should switch to the Browser tool.

Redirect Issues OpenClaw uses an SSRF guard. If a URL redirects to a private or internal IP address, the fetch will be blocked for security reasons.

If you need more help getting things running, check out the AI Setup Assistant.

What’s Next

• Perplexity Search setup
• Brave Search setup
• Browser tool documentation
• Firecrawl setup