OpenAI Deep Research Skill

This skill wraps OpenAI's Deep Research API so you can launch autonomous research runs (using o3-deep-research or o4-mini-deep-research) directly from Claude Code.

Setup

1. Dependencies: Install the OpenAI SDK.

   pip install openai python-dotenv
   

2. API Key Configuration: Export your OpenAI API key.

   echo "OPENAI_API_KEY=sk-..." >> .env
   if [ -f .gitignore ] && ! grep -q ".env" .gitignore; then echo ".env" >> .gitignore; fi
   

   The Deep Research endpoint requires access to the Early Access program. Ensure your account is enabled before calling the API.

Usage

Use scripts/run_deep_research.py to start a research job and stream the final report to stdout.

Command

python3 scripts/run_deep_research.py --prompt "<research_question>" [--model o4-mini-deep-research] [--effort medium] [--json]

Parameters

• --prompt (Required): The investigation prompt.
• --model (Optional): o4-mini-deep-research (default) or o3-deep-research.
• --effort (Optional): Reasoning depth, one of low, medium, high (default medium).
• --json (Optional): Emit the full response payload as JSON instead of just the synthesized report.
• --output (Optional): Write the report to a file path.

Example

python3 scripts/run_deep_research.py \
  --prompt "Assess the projected market impact of solid state batteries by 2030" \
  --model o3-deep-research \
  --effort high \
  --output reports/solid-state-batteries.md

Output

The script prints the synthesized Deep Research report (with citations) to stdout or saves it to the specified file. When --json is provided, the entire response payload (including intermediate steps, citations, and metadata) is emitted.

Features

• True Deep Research: Uses OpenAI's autonomous planning, browsing, and synthesis stack.
• Configurable Effort: Choose between faster runs (o4-mini-deep-research) or higher-quality runs (o3-deep-research).
• Intermediate Visibility: Optional JSON output exposes the tool traces and citations returned by the API.