Point-in-time Congress, BTC-ETF-flow, insider, token-launch and options-IV feeds; free + x402
Scored across 7 tools
Each tool maps to a distinct dataset (congress trades, ETF flows, insider buys, insider sells, token launches, launch wallet book, options IV), and the descriptions explicitly cross-reference their nearest neighbors (insider_buys vs insider_sells, launch_ledger vs launch_wallets) to reinforce the boundary. An agent can reliably select the right feed for a given question.
All seven names follow a uniform snake_case noun-based convention describing the dataset (congress_trades, etf_flows, insider_buys, insider_sells, launch_ledger, launch_wallets, options_iv). No mixing of verb styles or casing.
Seven read-only feeds is well-scoped for a data-feed server: each one represents a distinct source with no redundant or filler entries. Nothing feels padded or missing at the count level.
The feed surface covers a coherent set of alternative-data domains with rich metadata per record, and paired feeds (buys/sells, ledger/wallets) close obvious gaps. Minor absence of any discovery/listing or cross-feed query tool means an agent must know feeds by name, but this is workable.
US Congress stock transactions from official Senate ePTR and House Clerk disclosures (public domain). Every record carries happened_at (transaction date), disclosed_at (filing date) and observed_at (when this pipeline first saw it). Because the STOCK Act permits ~45 days between trade and disclosure, disclosed_at is the field to use for point-in-time work. Senate data is parsed to ticker level; House filings are metadata only.
Inputs: name, paid, since, ticker, +2 more.
Spot bitcoin ETF creation/redemption flows, computed as change in shares outstanding times NAV so price moves are removed and only genuine creations/redemptions remain. This series cannot be backfilled -- the issuer publishes only a current snapshot -- so every row was banked on the day it existed. span_days above 1 marks a row that spans a gap.
Inputs: paid, since, ticker, tickers.
SEC Form 4 open-market insider BUYS (public domain). Form-4 code P only: grants, option exercises, tax withholding and gifts are excluded because they are compensation events rather than conviction. An officer or director spending their own money is the signal this feed isolates. Carries observed_at (when recorded here) and happened_at (transaction date on the filing). For disposals use insider_sells. Since 2026-09-01 the feed also carries Schedule 13D/13G beneficial-ownership stakes (form=13D active intent, form=13G passive; direction=buy on an initial filing or a percent increase versus the last filing seen); filter with form=4 for Form 4 only.
Inputs: form, paid, since, ticker, +2 more.
SEC Form 4 open-market insider SELLS (public domain). Form-4 code S only. Read these with care: insiders sell for diversification, tax and pre-scheduled 10b5-1 plans, so a sale carries far less information than a purchase. Carries observed_at (when recorded here) and happened_at (transaction date on the filing). For acquisitions use insider_buys. Also carries Schedule 13D/13G stake DECREASES and exits below 5% (form=13D|13G, direction=sell versus the last filing seen); filter with form=4 for Form 4 only.
Inputs: form, paid, since, ticker, +2 more.
Robinhood Chain (chain id 4663) token launches that cleared a first-hour activity floor, with the early-buyer cohort: the first ~25 distinct wallets that bought (transaction sender, never the router), each with block delta from launch and quote spent. Every launch is graded at 24h and 7d (price multiple vs first buy, liquidity, holders). Launchpad is identified from the Uniswap v4 hook. Recorded forward from 2026-09-03 straight from chain logs: happened_at is the launch block time, observed_at is when the cohort was banked. Not backfilled.
Inputs: paid, quote, since, ticker, +5 more.
Wallet book derived from launch_ledger: every address seen in an early-buyer cohort, with how many launches it was early in, its 24h hit rate (launch traded at 2x or more vs its first buy), dead rate (no trades or below 0.2x), average block delta, launchpads used, tokens, and whether it also deployed launches. Repeat early buyers with a high hit rate are the cohort worth watching; this feed ranks, it does not recommend. Recomputed at each publish from the forward-only ledger.
Inputs: paid, since, ticker, wallet, +2 more.
Daily US equity option-chain structure per ticker, recorded forward from 2026-06-27 (~230 names, weekday close). Derived metrics only -- no quotes, strikes, greeks or spot are re-served: atm_iv_30d / atm_iv_90d (implied vol at the ~30d and ~90d expiries), iv_term (30d/90d; above 1 = backwardation/stress), skew_25d (25-delta put IV minus call IV; positive = put demand), iv_rank (percentile of today's 30d IV within THIS series' own history, with iv_hist_days saying how long that history is -- a rank over 10 days is not a rank), put/call ratios by open interest and by volume, net dealer gamma sign, max-OI call/put strikes as percent from spot, and an unusual-activity read (uoa_lean call/put/mixed from contracts with volume >= 3x open interest; EOD delayed data, no prints, so buyer vs seller is unknowable). happened_at is the trading day; observed_at is the day the chain was banked. Not backfilled; the source publishes only the current chain, so nobody can reconstruct these days.
Inputs: lean, paid, since, ticker, +4 more.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Paid market-data MCP: crypto, equities, smart-money, macro, wallet PnL. x402 per call on Base.
Live data for AI agents: crypto, Hyperliquid, stocks, news, macro, predictions. Free key or x402.
Verified market data for AI trading agents: quality-flagged candles, funding, OI, order flow. x402.
Crypto order flow, cross-venue derivatives and US market context for AI agents, with a track record.
Add one secure layer between your agents and this server.
