---
name: senpi-smart-money
description: >-
  Read where top Hyperliquid traders sit with the Senpi connector. Use for "what's smart money
  doing?", "are the whales long or short?", "smart money on ETH?", "big moves lately?", "who's
  buying?", "whale movements?", "what did smart money do today?". Calls get_smart_money_positioning (the cohort's
  current lean), get_smart_money_flow (what the cohorts did over a window) and get_whale_moves
  (top traders whose 4h P&L crossed a tier). Public wallets, read-only.
---

# Smart Money Radar

Three tools:

- `get_smart_money_positioning` — two cohorts: the **smart** cohort, the proven tier (lifetime
  realized ≥ $1M) that Senpi Signals' cards count, and a **crowd** sample (150 wallets with
  $10k–$100k realized); a holder is a position of at least $1,000. Returns `smart_leaning` — the
  coins where at least 5 proven traders lean past a net bias of 0.40 (`bias` is net/gross in
  −1..+1, with `members`, `n_long`, `n_short`, `net_usd`), strongest first — and `divergences`,
  where the two cohorts sit on opposite sides or 0.50 apart. Optional `asset` adds `smart` and
  `crowd` reads on that one coin, with the members' positions. `cohorts` names each cohort and its
  size; when the signals cohort can't be read, the smart cohort falls back to a 150-wallet sample
  of the ranking and its definition says so — mention that, since holder counts then differ from
  the asset card's.
- `get_smart_money_flow` — what the cohorts **did**, not where they sit: positions opened, added
  to, trimmed, closed and flipped over a `window` (`latest`, `1h`, `4h`, `24h`; default 1h), per
  asset and cohort (`proven`, `crowd`, `apex`, `whale`), with `net_usd` per asset and the
  `divergences` between cohorts; `include_events: true` adds the newest wallet moves of $10k or
  more behind the flow. Quote `age_s`: the read is a snapshot, and says how old. "Smart money is
  short ETH" is positioning; "smart money sold ETH this afternoon" is flow — use the right one.
  The per-asset totals are summed **across all cohorts**; one cohort's own moves come only from
  the events — pass `include_events: true` and group them by cohort before saying what the
  proven cohort, or the whales, did. The events are the newest 50 at most: when
  `events_cut_at_limit` is true they are a sample of the window, so say "among the latest moves",
  never a cohort's total for the window. Not every deployment offers this tool: when it is not in
  your tool list, answer a "what did they do" question with `get_whale_moves` (top traders whose
  4h P&L crossed a tier — not trades) and say the per-cohort flow is not available here.
- `get_whale_moves` — the 4h board's momentum events, one per trader (its newest that names
  `top_positions`), newest first, up to 8. Each is a top trader whose 4h P&L crossed a tier
  (`tier_label`, `delta_pnl`), with the positions they held — **not** an open or a close: never
  say a whale opened or closed anything from it. Optional `asset`, `window_hours` (default 4, the
  board's own window; max 168). When `read_cut_at_limit` is true the moves only reach back to
  `oldest_event_read`: say the span they cover, not the whole window.

## Presenting the result

- Write for the user, not about the tools. Never name a tool, argument or response field, never show
  a field as `key: value`, and never quote a raw code or enum value (snake_case labels, error codes,
  request ids): say in plain words what was read and what it found. A failed read is "couldn't be
  read right now", not its error. Wallet addresses, strategy ids and coin symbols may be shown.
- Lead with the lean: the top `smart_leaning` rows as "N proven traders, net X% short ETH", then
  any divergence — the proven cohort against the crowd is the core signal.
- An empty `smart_leaning` means no coin clears the 5-member / 0.40 bar right now — say that; it
  is an answer, not an error. With an `asset`, `small_sample: true` means fewer than 5 members
  hold it: report the positions, don't call it a lean.
- A cohort's read on one asset is thin when fewer than 20 of its wallets hold it, or one wallet
  holds more than half of its gross: say how many wallets that is and how much the largest holds,
  report the positions, and don't call that cohort's direction or compare it with the other
  cohort's. The crowd is a 150-wallet sample, so its read on one coin is often thin.
- `get_whale_moves` returns `UNAVAILABLE` when the board is down, never an empty list in its
  place.
- Attribute moves to addresses, not names, unless the result carries a name. These are public
  wallets; never speculate about identities.
- Quote `as_of`; positioning can be hours old by the time the user acts. No advice: describe the
  positioning, don't tell the user to copy it.

## Follow-ups to offer

- Depth on one trader → `get_wallet_profile` or `find_top_traders` (senpi-trader-research).
- Depth on one asset → `read_asset_signal`.
- The 4h board's ranking and where its gains concentrate, or one trader's positions behind a
  momentum event → `get_leaderboard` (`trader` for the wallet).