deer-flowlisted

# DeerFlow — Deep Research Agent DeerFlow is a LangGraph-based multi-agent system running at **http://localhost:8001**. It uses Claude Sonnet 4.6 as the primary model and Haiku 4.5 for fast sub-tasks. ## When to use - Deep multi-step web research (competitor analysis, market research, technical surveys) - Document analysis + synthesis across many sources - Complex tasks that benefit from parallel sub-agent decomposition - Generating structured reports, summaries, or slide decks from research ## Prerequisites The DeerFlow backend must be running. Start it in a separate terminal: ```bat cd "C:\Users\A\Documents\CURSOR PROJECTS\deer-flow" start.bat REM Reads your Claude Max token from ~/.claude/.credentials.json automatically REM Backend starts at http://localhost:8001 ``` No manual API key needed — it reads your Claude Max subscription token from `C:\Users\A\.claude\.credentials.json` automatically. ## Check if running ```bash curl -s http://localhost:8001/health || echo "DeerFlow not running" ``` ## API usage (call from Claude Code) ### Send a research task ```python import httpx, json BASE = "http://localhost:8001" THREAD_ID = "research-session-1" # Create or continue a thread resp = httpx.post(f"{BASE}/api/langgraph/threads", json={"thread_id": THREAD_ID}) # Run a task resp = httpx.post( f"{BASE}/api/langgraph/threads/{THREAD_ID}/runs", json={ "assistant_id": "default", "input": {"messages": [{"role": "user", "content": "YOUR TASK HERE"