cookbook — working agent recipes

copy-paste these into Claude Code / Cursor / Cline / any MCP client. 13 tools (3 free / 10 Pro $19/mo), no signup for free tier.

Use macro_tape. Tell me the macro regime in 3 sentences. Cite specific
numbers — last price, 1d return, 5d return. Then list 2 tickers from
sec8k_material_today that might benefit from this regime. Be terse.

Run insider_buy_clusters (min_insiders=3) and sec8k_material_today
(items=["2.02","8.01","5.02"]). Find any ticker that appears in both
within the last 24h. For each match, summarize what the insiders bought
and what the 8-K said. End with the bear case in one sentence.

Use confluence_today with min_signals=2. For each asset/ticker, explain
in 2 sentences why these signals stacking is meaningful. Skip obvious
noise (DeFi LP pools with stablecoins). Highlight the most contrarian
stack and what would invalidate it.

Use hl_vault_leaderboard with sort_by=score, limit=5. For each, flag any
vault with high return but high drawdown — that's noise, not signal. Pick
2 you'd actually copy-trade and explain why composite_score isn't enough
to trust them blindly.

Use top_yield_farms with min_apy=15. Strip any pool where realistic_apy
above 500% (probably broken data) or TVL below $1M (illiquid). For
survivors, rank by realistic_apy / sqrt(TVL) as crowded-vs-uncrowded
proxy. Recommend 3 with reasonable APY-to-risk profile.

Use etf_flow_today and macro_tape. Check if BTC ETF inflow direction
matches BTC's 5d return direction. If they diverge (ETF inflow but BTC
down 5d), that's a leading signal. Tell me which way and what it might
mean for 1-2 week horizon.

Use active_airdrop_farms with min_tvl_usd=5000000 and min_apy=10.
gap_apy is what you earn while waiting; the airdrop itself is bonus,
not promised. Pick 3 with cleanest setup — positive gap_apy means the
farm pays you to wait.

Use polymarket_whale_positions with min_position_usd=50000. For each
whale, calculate win_rate × position_size_USD as conviction score. List
top 3 by conviction and tell me which crypto-related markets they're
betting on. Don't blindly copy — assess each market thesis first.

build your own bot

each recipe is a starting point. for production use, combine multiple recipes into a Python script that runs on a cron. example template:

import urllib.request, json

def fetch(path):
    req = urllib.request.Request(f"https://falsifylab.com{path}")
    with urllib.request.urlopen(req, timeout=20) as r:
        return json.loads(r.read())

# combine confluence + macro for daily research
conf = fetch("/api/confluence?min_signals=2")
macro = fetch("/api/macro")

# your logic here
print(conf, macro)

or fork the agent starter repo — pre-wired Cline/Cursor config + 12 example prompts.

get more

Pro $19/mo — real-time + 100 results + 90d history → try live demo → earn 30-40% recurring →