webresearcher
v1.0.0
Published
CLI tool that uses GPT-5.2 with web search for up-to-date research queries
Downloads
11
Maintainers
Readme
webresearcher
A CLI tool that uses GPT-5.2 with web search for up-to-date research queries.
Why?
Claude code's built-in web search is great for quick fact-lookups, but bad at detailed research.
Evaluated against Claude's built-in WebSearch tool across 14 query types. webresearcher won 13/14 on quality while remaining cost-effective (~$0.03/query).
Installation
Set your OpenAI API key in your path:
export OPENAI_API_KEY="your-key-here"Requires Bun runtime.
# Clone the repo
git clone https://github.com/sky-xo/webresearcher.git
cd webresearcher
# Install dependencies
bun install
# Link globally
bun linkUsage
# Basic query (uses --effort low by default)
webresearcher "how do I stream responses with Vercel AI SDK?"
# Complex troubleshooting (use medium effort)
webresearcher --effort medium "why might Next.js build fail with module not found for an installed package?"
# See token usage and cost
webresearcher --verbose "current Bitcoin price"Effort Levels
| Effort | Avg Latency | Avg Cost | Best For | |--------|-------------|----------|----------| | low (default) | ~17s | ~$0.03 | Most queries | | medium | ~29s | ~$0.04 | Complex troubleshooting | | high | ~60s+ | ~$0.08+ | Critical decisions |
Claude Code Integration
This tool was built to be used as a Claude Code skill. Install the skill:
webresearcher --install-skillThis copies the skill definition to ~/.claude/skills/webresearcher/SKILL.md. Claude will then know when to use webresearcher vs the built-in WebSearch tool.
Quick trigger: Tell Claude to "use wr" and it will use this tool.
When to Use What
| Query Type | Recommended Tool | |------------|------------------| | Simple factual ("latest version of X") | WebSearch | | Comparisons | webresearcher | | How-it-works explanations | webresearcher | | Best practices | webresearcher | | Troubleshooting | webresearcher --effort medium | | Documentation/code examples | webresearcher |
Flags
| Flag | Description |
|------|-------------|
| --effort low\|medium\|high | Reasoning effort (default: low) |
| --verbose | Show token usage and cost |
| --install-skill | Install Claude Code skill |
License
MIT
