datasets

every tokenized stock priced three ways every five minutes: the 24/7 dex, the live us-market print (only while the session was open and fresh), and the chainlink oracle — with the premium and drift computed at capture, under the session regime the row was captured in.

• token_address + ts — pk, one row per stock per 5-min capture
• dex_price_usd / equity_price_usd / oracle_price_usd — the three legs; a leg that was not fresh at capture is null, never carried forward
• premium_pct — dex vs live print (session open only) · drift_pct — dex vs frozen oracle
• market_state — the session regime the row was captured under

curl -s https://rhxbt.com/api/v2/stocks/premiums | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_stock_premiums","arguments":{}}}'

erc-8056 multiplier transitions — splits and stock dividends — detected from the transfer ratio at a specific block (chain-dated facts), plus the forward calendar read from each contract's newUIMultiplier()/effectiveAt() before the change applies.

• token_address, previous_multiplier → new_multiplier, ratio (>1 reads as a forward split)
• observed_at + observed_block — the chain timestamp of the first transfer at the new ratio
• unique (token, block) — re-ingestion records a transition once
• companion table pending_corporate_actions: scheduled ratio, effective_at, first_seen_at (the announcement clock)

curl -s https://rhxbt.com/api/v2/stocks/corporate-actions | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"pending_corporate_actions","arguments":{}}}'

one row per stock per utc day: transfer counts with the off-hours share (the 24/7 settlement read), and primary-market mint/burn in shares at the multiplier in force at each block. the long-run record behind the settlement heatmaps.

• day + token_address — pk, utc-midnight bucket
• transfers / off_hours — activity and how much of it happened while the us session was closed
• minted / burned — shares tokenized and redeemed, multiplier-adjusted
• last_ts — max(ts) inside the bucket: a bucket cannot claim to be fresher than its newest row

curl -s https://rhxbt.com/api/v2/stocks/flows | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_equity_flows","arguments":{}}}'

delivery-versus-payment prints: a stock leg and a cash leg in one transaction between the same two parties in opposite directions — real executions at real size that never touched a pool, invisible to every dex-based index.

• tx_hash — pk · token_address, cash_token, buyer, seller
• shares (multiplier-adjusted) + cash_amount → implied_price_usd — an execution price, not a quote
• during_session — whether the us regular session was open at the block
• block, ts — chain-dated

curl -s https://rhxbt.com/api/v2/stocks/graph | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_settlement_graph","arguments":{}}}'

lighter perpetuals (the chain's native perp venue, usdg-quoted): one current row per market, and the hourly funding series — one bucket per settlement hour, append-only, a failed pass leaves a gap and never a zero. where the ticker is a tokenized stock we track, the read model joins our dex leg for the perp-vs-spot basis nobody else can compute.

• perp_markets: market_id, symbol, mark/index price, open interest (base + usd), 24h volume, hourly funding_rate, next_funding_ts, observed_at (the collector's clock — the venue publishes none)
• perp_funding_history: market_id + ts (hour bucket) — pk; funding_rate, mark, index, observed_at

curl -s https://rhxbt.com/api/v2/perps | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"perps_markets","arguments":{}}}'

totalSupply() per tracked stablecoin, hourly, counted at par — plus the issuance tape (stable_flows: every mint from and burn to the zero address, block-timestamped). the float behind every other board; says nothing about peg.

• token_address + ts — pk, hourly reading
• total_supply (normalized by on-chain decimals) · raw_supply (the raw uint256 as a string, so a decimals mistake is recoverable)
• holders — from blockscout counters where reported, else null
• companion table stable_flows: mint/burn tape with counterparty, block, ts

curl -s https://rhxbt.com/api/v2/stables | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"stablecoin_flows","arguments":{}}}'

sequencer economics measured from blocks, one row per settled utc hour from a bounded, evenly spaced sample: transactions, gas, l2 fees actually paid (gasUsed × effectiveGasPrice from receipts), contract deploys, and erc-4337 entrypoint traffic — the sponsored population the gas-subsidy cliff applies to.

• ts — pk, hour bucket start · txs, gas_used, deploys, aa_txs
• fees_wei — decimal string (wei exceeds float precision); null when no sampled receipt carried a price — absent, never zero
• sampled + sampled_blocks/total_blocks — load-bearing honesty: extrapolated figures say so all the way to the page
• a missing hour is a missed pass, never a zero row

curl -s https://rhxbt.com/api/v2/chain/economics | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"chain_economics","arguments":{}}}'

every chainlink feed on the chain with a verdict every five minutes: ok | stale (no new round past the feed's own observed cadence while the session is open) | paused (the stock contract's advisory oraclePaused()) | diverged (vs the live print or the dex past the broken-feed bar).

• feed_address + observed_at — pk · symbol, linked token_address where the feed backs a stock
• round_id, price_usd, last_updated_at, stale_sec · expected_sec — the median inter-update gap over 24h, because the registry heartbeat is a 24h maximum
• paused — null when the contract did not answer (unknown, never assumed ok)
• dev_vs_dex_pct / dev_vs_equity_pct → status

curl -s https://rhxbt.com/api/v2/oracle-health | jq .

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"chainlink_feed","arguments":{}}}'

every tracked caller's token calls with verified onchain outcomes: entry price at first observation after the tweet, peak, the anti-wick 24h-later price (a win must still be a win a day later), and a rugged flag when the shilled pool later collapses to dust. the evidence behind the kol leaderboard.

• author_key + cand_key — pk · author_handle, token_address, tweet_id, tweet_ts
• entry_price/mcap/liquidity + entry_ts — first observation at-or-after the tweet; null until the radar sees it
• peak_price + peak_ts · price_at_24h — the anti-wick win metric · rugged_at
• verified — token confirmed to exist onchain

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_kol_leaderboard","arguments":{}}}'

curl -s https://rhxbt.com/mcp -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_kol_leaderboard","arguments":{}}}'

the wallet pnl leaderboards, precomputed on the agent because the 7-day fold over the full swap tape is a scan no synchronous render path may run (one such request was once a full site outage). every row's pnl carries valuation provenance — a wallet with undatable marks is excluded, not guessed.

• kv payload: { computedAt, validUntil, rows } — rows are eoa-only (contracts filtered by bytecode check)
• per wallet: realized/unrealized pnl with valuation clocks, trade counts, low-sample flag
• reads fail closed past validUntil — a stale board is withheld, never re-served

# no public rest route yet — served on https://rhxbt.com/wallets

# no mcp tool yet — the boards render at https://rhxbt.com/wallets