The web, from your terminal.

Search the web, extract pages, get grounded answers, and run deep research with one command: web. Choose your providers, bring your API keys, and use the same configuration in your terminal, TypeScript applications, and pi.

pi remove npm:pi-web-providers
pi install npm:webfox

• Terminal: Install the web command with Homebrew:

  brew install mavam/tap/webfox

  Or use npm:

  npm install -g webfox

• Nix: Run the CLI directly:

  nix run github:mavam/webfox -- --help

  The flake exports packages.<system>.default for Linux (aarch64, x86_64) and macOS (Apple Silicon).

• TypeScript: Add webfox as an application dependency.

• Pi: Install the extension:

  pi install npm:webfox

  This installs the extension, not the web command on your shell's PATH. Then follow Use with Pi to select a web provider and supply its API key.

Set an API key and make your first request. For example, with Brave Search:

export BRAVE_SEARCH_API_KEY=…
web search "Node.js release notes" --provider brave

Save your provider choice so you don't need to repeat it:

web config default search brave
web search "TypeBox validation"

Choose a provider for each capability, or override it with --provider. API keys alone never select a provider.

web search "Node.js cancellation" "Bun cancellation"
web contents https://example.com/a https://example.com/b --provider tavily
web answer "What is MCP?" --provider openai --model gpt-6-astra
web research "Compare databases for an analytics service" --provider gemini --timeout 20m

Each provider needs its own credentials. Research runs in the foreground and shows progress while you wait. Ctrl-C stops waiting, but may not cancel an already-running, billable research job at the provider.

web providers
web providers openai
web search --help
web search --provider openai --help

Provider-specific help lists the options available for that provider, such as models, source filters, and page extraction settings.

Quote each independent query or question. Search and answer accept up to ten inputs; research accepts one brief. Omit the input to read from a pipe or file:

# Read from files
web search < query.txt
web answer --provider openai < question.txt
web research --provider gemini < brief.md
web contents --provider tavily < urls.txt

# Pipe input
echo "What is MCP?" | web answer --provider openai
echo "https://example.com" | web contents --provider tavily

# Save or process JSON results
web search "TypeBox" --format json > results.json
web search "TypeBox" --format json | jq -r '.results[] | select(.ok) | .value.results[].url'

Contents reads one URL per line. The other commands read the entire stdin stream as one input. Positional arguments take precedence over stdin; - is optional. Text is the default, including when piped. Use --format json for structured results; the last example uses jq to extract result URLs.

Results go to stdout; progress and errors go to stderr. --quiet hides progress and success notices, not errors. Use --timeout 30s or --timeout 20m to set a deadline, and --no-color or NO_COLOR to disable colors.

See the CLI reference for output formats, exit codes, and scripting details.

Use different providers for different tasks:

See the provider guide for credentials, examples, and caveats.

Save choices with web config default, or edit YAML for more control:

web config path
web config show
web config validate

show hides credentials. validate checks your settings without making requests. The default file is ~/.config/webfox/config.yaml, respecting XDG_CONFIG_HOME or, on Windows, APPDATA. Override it with WEBFOX_CONFIG or --config <path>.

For example:

$schema: https://unpkg.com/webfox@latest/dist/config.schema.json
defaults:
  search:
    provider: brave
    maxResults: 5
  answer:
    provider: openai
providers:
  openai:
    options:
      answer:
        model: gpt-6-astra

Standard API key environment variables work without a credentials section. Run web providers <id> to find the names for your provider. You can also read keys from a password manager rather than store them in the file.

See the example configuration and configuration reference for credential commands, timeouts, retries, and provider defaults.

Install the Pi extension:

pi install npm:webfox

To also use the web command in your terminal, follow Installation. Both installations share the same Webfox configuration.

Add to ~/.config/webfox/config.yaml:

defaults:
  search:
    provider: brave

Start Pi with your API key:

export BRAVE_SEARCH_API_KEY=…
pi

Ask Pi: Search the web for the latest Node.js release notes.

Add contents, answer, or research defaults for more tools. Restart Pi after changing defaults.

import { createWebfox } from "webfox";

const web = createWebfox();
const document = await web.search({
  provider: "brave",
  queries: ["Node.js AbortSignal"],
  maxResults: 5,
});

for (const result of document.results) {
  if (result.ok) console.log(result.value.results);
  else console.error(result.error.message);
}

The client also provides contents, answer, and research. See the library reference for request controls, configuration, progress, and errors.

• No provider selected: Pass --provider <id> or save a default with web config default <capability> <provider>.
• Missing credentials: Run web providers <id> to check the required key names, or see the provider guide.
• Invalid configuration in pi: Pi continues to start, but Webfox registers no tools and reports the error (on stderr in print and JSON modes). Fix the configuration, then restart Pi or run /reload to load the extension again.
• No pi tools: Select default providers in the shared configuration and restart pi. Installing the extension or setting keys alone isn't enough.

MIT