ChatGPT Deep Research

# ChatGPT Deep Research Use ChatGPT's Deep Research feature to conduct thorough research on any topic. Deep Research browses the web extensively and produces a detailed report with sources. ## Prerequisites - Chrome MCP browser tools available (mcp__claude-in-chrome__*) - Logged in to chatgpt.com (James's account is pre-authenticated, Pro plan required) ## Step 1: Navigate to Deep Research ``` 1. Create a new tab or use existing: mcp__claude-in-chrome__tabs_create_mcp 2. Navigate to: https://chatgpt.com/deep-research 3. Read page to confirm — look for "Deep research" label and the text input ``` The Deep Research page has: - A text input with placeholder "Get a detailed report" - A "Sites" button to scope research to specific sites - Previous reports listed at the bottom ## Step 2: Submit a Research Query ``` 1. Read page with filter: interactive 2. Find the textbox labeled "Chat with ChatGPT" (there are two — use the one with placeholder "Get a detailed report") 3. Use form_input to type your research query 4. Click the "Send prompt" button to submit ``` **Writing effective research queries:** - Be specific about what you want to learn - Include scope constraints (time period, geography, industry) - Ask for comparisons if relevant - Examples: - "Research the top 10 AI agent frameworks in 2025, compare their architectures, community size, and production readiness" - "Find all competitors to Lovable.dev for AI-powered app building, their pricing, features, and user reviews" - "Research the current state of browser automation tools for AI agents, including MCP-based solutions" ## Step 3: Monitor Progress Deep Research takes 3-15 minutes. It shows progress as it browses. ``` 1. Wait 30 seconds, then read the page 2. Look for progress indicators — it shows sites being visited and thinking steps 3. Do NOT interact with the page while research is running 4. Check every 30-60 seconds until you see the full report 5. The report is complete when you see source citations and a structured response ``` **Important:** Deep Research can take up to 15 minutes. Be patient. Do not refresh or navigate away. ## Browser Automation Tips ### Research Confirmation Step After submitting a query, ChatGPT shows a research plan with steps and a "Start" button with a countdown timer. The research auto-starts when the timer reaches zero (~50 seconds). To start immediately: ``` 1. Wait 2-3 seconds for the plan to appear 2. The "Start" button is NOT a standard button element — use JavaScript: document.querySelector('button') matching text "Start" — or just wait for auto-start 3. The countdown is visible next to the Start button (e.g., "Start 50", "Start 24") ``` ### Text Input The chat input is a contentEditable div, not a textarea. Use click + type: ``` 1. Click on the input area (the "Get a detailed report" placeholder) 2. Use computer action "type" to enter your query 3. Click the send button (blue circle arrow) — standard click works here ``` ### Developer Mode ChatGPT may show "Developer mode" and "Memory is not used for this chat" on the Deep Research page. This is normal — deep research runs without memory. ## Step 4: Extract the Results Once the report is complete: ``` 1. Use get_page_text to extract the full text content of the report 2. The report includes: - Executive summary - Detailed findings organized by topic - Source citations with URLs - Sometimes tables or comparisons 3. Save the report text for further processing ``` ## Step 5: Save and Reference ``` 1. The research URL will be: chatgpt.com/c/<conversation-id> 2. Save this URL for future reference 3. Previous reports are listed on the Deep Research landing page 4. You can continue the conversation to ask follow-up questions ``` ## Using Sites Filter To scope research to specific websites: ``` 1. Before submitting your query, click the "Sites" button 2. Add specific domains to search (e.g., github.com, arxiv.org) 3. This focuses the research on authoritative sources ``` ## Retrieving Previous Reports ``` 1. Navigate to: https://chatgpt.com/deep-research 2. Scroll down to see the "Reports" section 3. Click on a previous report to view it 4. Use get_page_text to extract the content ``` ## Tips - Deep Research is best for complex, multi-faceted questions that need web browsing - For simple factual questions, use regular ChatGPT instead - The feature is rate-limited — Pro plan gets ~10 deep research queries per month - Results include source URLs which can be verified independently - You can ask follow-up questions in the same conversation to dig deeper ## Running on Mac Minis These instructions can also be executed on the Mac Minis (bots, vault, jarvis) using the mac-control MCP server instead of chrome-in-chrome. ### Via mac-control MCP Server (Port 8200) The mac-control server on each Mac exposes the same GUI automation primitives: ```bash # Take a screenshot to see the screen curl -s http://bots.local:8200/tools/screenshot # Open a URL in Chrome curl -s http://bots.local:8200/tools/open_url -d '{"url": "https://chatgpt.com"}' # Click at screen coordinates curl -s http://bots.local:8200/tools/click -d '{"x": 500, "y": 300}' # Type text curl -s http://bots.local:8200/tools/type_text -d '{"text": "Research the latest advances in..."}' # Press keyboard shortcut curl -s http://bots.local:8200/tools/key_press -d '{"keys": "return"}' # Run AppleScript curl -s http://bots.local:8200/tools/applescript -d '{"script": "tell app \\"Google Chrome\\" to activate"}' # Find text on screen (OCR) curl -s http://bots.local:8200/tools/find_on_screen -d '{"text": "Submit"}' ``` ### Via Chrome DevTools Protocol (Port 9222) For direct browser control without GUI coordinates: ```bash # List open tabs curl -s http://bots.local:9222/json # Navigate a tab to a URL curl -s http://bots.local:9222/json/new?https://chatgpt.com # Execute JavaScript in a tab (via WebSocket — use a CDP client) ``` ### Via SSH + cliclick (Simple Fallback) ```bash # Open URL ssh bots 'open https://chatgpt.com' # Wait for page load sleep 3 # Screenshot + analyze ssh bots 'screencapture /tmp/screen.png' scp bots:/tmp/screen.png ./screen.png # Click at coordinates ssh bots '/opt/homebrew/bin/cliclick c:500,300' # Type text ssh bots '/opt/homebrew/bin/cliclick t:"Research the latest advances in..."' # Submit ssh bots '/opt/homebrew/bin/cliclick kp:return' ``` ### Recommended Machine Use **bots** (192.168.86.50) for browser automation — it has full GUI support, screencapture works, and Chrome with remote debugging is running.