How to Install and Use Brave Search in OpenClaw

The brave-search skill is a standalone ClawHub package with its own BRAVE_API_KEY — it does not share credentials with any Tavily skills you may have installed. This means you get a completely separate API quota and a genuinely independent search backend under a single OpenClaw chat interface. Open your terminal and run the install command. ClawHub downloads the skill package from the registry, validates it, and registers it with your local OpenClaw installation. You will see the confirmation message listing the required configuration key: BRAVE_API_KEY. If you already have other search skills installed (tavily-ai-search, tavily-web-search, web-search-by-exa), brave-search installs alongside them without conflict. OpenClaw will have multiple search backends available and can route to the appropriate one based on your prompt context or explicit instructions.

1# Install Brave Search skill
2clawhub install brave-search
3
4# Confirm installation
5clawhub list | grep brave
6
7# See all available configuration options
8clawhub info brave-search

Navigate to api.search.brave.com and sign up for a developer account. Brave's API is separate from the Brave browser account — you can have one without the other. Click 'Subscribe' on the API product page to see available plans. Brave offers a free tier ('Data for Independent Developers') with 2,000 queries per month at no cost. This is the most generous free tier among the major search skills available in OpenClaw, and it covers a useful volume for individual users. Paid plans start at $5/month for 20,000 queries. Once you have subscribed to a plan (the free tier is sufficient to start), navigate to 'API Keys' in your Brave API dashboard and click 'Generate Key'. Copy the API key — it is a long alphanumeric string. Unlike Tavily keys, Brave API keys do not have a distinguishing prefix, so the string may look like a generic token. Keep it safe — you can generate a new one from the dashboard at any time. Note that Brave's API key is specifically for their Search API. It is not the same as a Brave browser sync key or Brave Rewards key.

Set your Brave API key in OpenClaw's configuration. Unlike the Tavily skills, Brave Search uses its own environment variable name: BRAVE_API_KEY. Both variables can coexist in your OpenClaw .env file without conflict — each skill reads only the variable it expects. Use the clawhub config command to set the key, or add it directly to your ~/.openclaw/.env file. After setting it, reload the OpenClaw configuration to activate the skill. The variable name must match exactly: BRAVE_API_KEY (uppercase, underscore between the words). A common mistake is using BRAVE-API-KEY (hyphens) or BRAVE_SEARCH_KEY (different name), both of which cause the skill to fail authentication silently.

1# Set Brave API key (separate from TAVILY_API_KEY)
2clawhub config set BRAVE_API_KEY your-brave-api-key-here
3
4# Or add to ~/.openclaw/.env alongside other keys
5# ~/.openclaw/.env
6TAVILY_API_KEY=tvly-your-tavily-key
7BRAVE_API_KEY=your-brave-api-key
8
9# Reload to activate
10clawhub reload
11
12# Verify
13clawhub config get BRAVE_API_KEY

Open OpenClaw chat and run a test query. To verify that OpenClaw is using Brave's independent index rather than a Google-backed skill, you can ask a query on a topic where you know Brave might surface different sources — privacy advocacy sites, independent tech blogs, or international content. Brave Search results include titles, URLs, content snippets, and metadata such as published date and page age. The response format is similar to Tavily Web Search but with Brave's own ranking signal applied — which favors independent, non-sponsored sources and weights editorial quality over SEO optimization. This makes Brave's ranking noticeably different from Google-derived results for many queries. If OpenClaw has multiple search skills installed, explicitly mention 'use Brave' or 'search with Brave' in your prompt to ensure it routes to this skill rather than defaulting to Tavily.

Brave Search supports configuration options for result count, safe search level, and freshness filtering. For most users the defaults work well, but teams running OpenClaw in professional contexts — especially where content safety matters or where stale content is a problem — benefit from explicit configuration. You can also set up OpenClaw to automatically route certain query types to Brave Search. Privacy-sensitive queries (anything involving personal names, business intelligence, competitive analysis) are natural candidates for Brave. General research queries where synthesis quality matters are better routed to Tavily AI Search. This multi-skill routing strategy lets you get the best of multiple independent indexes. RapidDev can advise teams on advanced multi-skill routing configurations in OpenClaw for research and intelligence workflows that require comprehensive cross-source coverage.

1# ~/.openclaw/skills/brave-search.yaml
2brave-search:
3  count: 10              # results per query (default: 10, max: 20)
4  safesearch: moderate   # off | moderate | strict (default: moderate)
5  freshness: pw          # pw=past week, pm=past month, py=past year (optional)
6  search_lang: en        # language code for results (optional)
7  country: us            # country code for results localization (optional)
8  extra_snippets: true   # include additional content snippets per result