Claude Web Research Task

Use Claude.ai's Research feature from Claude Code via Chrome automation

This setup lets you spawn a background agent that opens Claude.ai, toggles the Research feature, runs your query, waits for completion, and returns the results, all while you continue working.

Why This Exists

Claude Code doesn't have native access to Claude.ai's Research toggle. But Claude in Chrome can control the browser. This bridges the gap by having Claude Code spawn a background agent that automates the web interface.

Key insight: Research takes 5 to 30+ minutes, so this MUST run as a background agent. The task feature ensures it won't be skipped or inlined into other workflows.

Prerequisites

Requirement	Details
Claude Code	v2.0.60+ (background agents support)
Claude Plan	Pro, Team, Max, or Enterprise
Claude in Chrome	Extension installed and authenticated
Chrome	Running with extension active

Quick Start

1. Install the Skill

chmod +x setup.sh
./setup.sh

Or manually copy files to ~/.claude/skills/web-research-task/

2. Add to Your Project's CLAUDE.md

Copy the contents of CLAUDE-snippet.md into your project's CLAUDE.md file.

3. Start Claude Code with Chrome

claude --chrome

4. Run Research as Background Agent

Spawn a background agent to run web research for: "Latest Laravel 12 features and breaking changes from previous versions"

Or use the slash command:

/research-via-web "Your query here"

Important: Press Ctrl+Shift+B if not automatically backgrounded.

File Structure

claude-web-research-task/
├── SKILL.md                    # Main skill documentation
├── agent-prompt.md             # Instructions for the background agent
├── CLAUDE-snippet.md           # Copy this to your project's CLAUDE.md
├── setup.sh                    # Installation script
├── README.md                   # You are here
└── commands/
    └── research-via-web.md     # Slash command definition

How It Works

┌─────────────────┐
│  Claude Code    │
│  (Main Session) │
└────────┬────────┘
         │ Spawn background agent
         ▼
┌─────────────────┐
│ Background      │
│ Agent           │
└────────┬────────┘
         │ Uses claude --chrome
         ▼
┌─────────────────┐
│ Claude in       │
│ Chrome          │
└────────┬────────┘
         │ Navigates to
         ▼
┌─────────────────┐
│ claude.ai       │
│ Research Mode   │
└────────┬────────┘
         │ Runs query (5-30 min)
         ▼
┌─────────────────┐
│ Results saved   │
│ to ./docs/      │
└─────────────────┘

Output

Results are saved as Markdown files:

Location: ./docs/research/

Filename: YYYYMMDD-HHMMSS-query-slug.md

Format:

# Research: [Query Title]

**Generated**: 2026-01-26T14:30:00Z
**Duration**: ~15 minutes
**Source**: Claude.ai Research Feature

---

[Full research content with citations]

---

## Sources Referenced
- Source 1
- Source 2
...

Checking Status

While the background agent runs:

Show status of background agents

Or type /bashes in Claude Code.

Troubleshooting

Problem	Solution
"Chrome not connected"	Run claude --chrome to start with Chrome integration
"Research toggle not found"	Ensure you have a paid Claude plan
"Login required"	Open Claude in Chrome sidebar and authenticate first
"Task was skipped"	Ensure you're running as background agent (Ctrl+Shift+B)
"Timeout"	Research can take 30+ min; don't interrupt

Tips for Your Workflow

1. Batch research requests: Queue multiple research queries as separate background agents
2. Check periodically: Use /bashes to see progress without interrupting
3. Use specific queries: More specific queries get better Research results
4. Review citations: Research output includes sources; verify important claims

Combining with Your Agent Setup

Since you have 13 specialized agents for Laravel/Vue development, consider:

1. Research Agent: Add this as a 14th specialized agent type
2. Pre-task research: Run research before major refactoring or migrations
3. Documentation: Use research results to inform your docs agent

Example workflow:

1. /research-via-web "Strangler Fig pattern best practices for Laravel monolith migration"
2. [Continue other work while research runs]
3. [Research completes, saved to ./docs/research/]
4. @docs/research/strangler-fig.md Now implement phase 1 of migration

Credits

Original idea by @EleanorKonik (eleanorkonik.com). This repo formalizes her concept into a working implementation.

License

MIT. Use freely in your projects.