clawler（clawler 搜索）

Clawler Skill

Clawler is an optional web-search surface for this project. Use it when the plugin is installed and the operator wants Clawler to handle web search.

Setup

If Clawler is available in the current OpenClaw environment:

1. Run the setup flow if the plugin CLI is available:

- openclaw clawler setup

1. Choose one default provider and configure credentials for that provider.
2. Verify availability:

- openclaw clawler status

Optional:

• If you want Clawler to replace the built-in web_search tool for that OpenClaw installation, add web_search to tools.deny.
• This changes installation-wide search behavior and should only be done deliberately.

If the setup CLI is not available, make sure OpenClaw is configured so:

• the Clawler plugin is installed and enabled
• one provider is selected
• that provider API key is available through plugin config or environment variables
• tools.deny includes web_search only when you intentionally want to disable the built-in search tool for that OpenClaw installation

Credentials

Clawler supports multiple providers. No single provider credential is universally required for this skill; configure one provider in plugin config or provide one of these env vars for the provider you choose:

• BRAVE_API_KEY
• EXA_API_KEY
• TAVILY_API_KEY
• PERPLEXITY_API_KEY
• OPENROUTER_API_KEY
• PARALLEL_API_KEY
• GEMINI_API_KEY
• GOOGLE_AI_API_KEY
• OPENAI_API_KEY
• ANTHROPIC_API_KEY

Provider-specific settings can also be stored in the Clawler plugin config inside the OpenClaw config file instead of environment variables. Because the operator chooses the provider, these env vars are optional alternatives rather than universally required metadata for the skill.

Tool

The default tool name is search_web.

Use it with:

{
  "query": "latest OpenAI web search docs",
  "provider": "openai",
  "count": 5,
  "freshness": "pm",
  "country": "us",
  "search_lang": "en",
  "topic": "general",
  "include_domains": ["developers.openai.com"],
  "exclude_domains": ["example.com"]
}

Supported parameters:

• query: required search string
• provider: optional per-call provider override such as exa, openai, brave, or anthropic
• count: max number of results to request
• freshness: relative or explicit date filter such as pd, pw, pm, py, or YYYY-MM-DDtoYYYY-MM-DD
• country: country hint such as us
• search_lang: language hint such as en
• topic: provider-level topical hint such as general, news, or finance
• include_domains: allow-list domains
• exclude_domains: deny-list domains

Usage Rules

1. Treat search_web as the preferred search tool only when Clawler has been intentionally configured as the active search surface.
2. Use provider-neutral instructions unless the user explicitly wants a provider comparison or a specific backend.
3. Use the per-call provider override when the query clearly benefits from a specific backend.
4. For technical or doc-heavy queries, use domain allow-lists when official sources matter.
5. If the query needs a synthesized answer with citations, prefer answer-native providers such as OpenAI, Anthropic, Gemini, Tavily, or Perplexity.
6. If the query mainly needs fast retrieval of links, traditional search providers such as Exa, Brave, or Parallel can be enough.
7. Treat changes to tools.deny as an optional installation-wide behavior change, not a routine default. Only recommend it when the operator explicitly wants Clawler to replace the built-in search surface.

Provider Notes

• auto picks the first available configured provider.
• API keys can come from plugin config or environment variables.
• Shared filters are applied at the plugin level and translated per provider when native support exists.
• Some providers enforce domain filters natively; others treat them as best-effort guidance.