PoofBit

npx add-mcp 'https://poofbit.com/mcp'

GDPR Compliance

We use cookies to ensure you get the best experience on our website. By continuing to use our site, you accept our use of cookies (and chaos).

PoofBit API

Use the Poof Chaos Index and the Poof Index in your bots, dashboards and AI agents.

The Poof Chaos Index (PCI) is a market-wide, direction-neutral measure of turbulence from 0 to 100, computed from public market data. It is designed to be read next to Fear & Greed: sentiment tells you which way the crowd leans, PCI tells you how violently the market is moving.

The Poof Index assesses single coins from 11 to 111: the higher, the deeper the coin is in a poof state. Everything is available as JSON over HTTPS and as tools for AI agents over MCP.

Get an API key

Free and instant. Every request to /api/* and /mcp needs a key. Need higher limits? Create an account to get account tokens (120/min) and a watchlist.

Quick start

Plain JSON over HTTPS. Put your key in an environment variable and send it as a Bearer token:

export POOFBIT_KEY=pb_your_key_here
curl -H "Authorization: Bearer $POOFBIT_KEY" https://poofbit.com/api/pci

Authentication

Send your key with every request, in either header:

• Authorization: Bearer YOUR_KEY (preferred)
• X-API-Key: YOUR_KEY

Keys are never accepted in the URL, because URLs end up in logs and browser history. We store only a hash of your key, so a lost key can't be recovered: create a new one.

A missing, unknown or revoked key returns 401 with a WWW-Authenticate: Bearer header:

{ "error": "Missing or invalid API key", "docs": "https://poofbit.com/docs#authentication" }

Going over the rate limit returns 429.

Free vs account

Tier	Where to get it	Limits
free	The form above, no account	30 requests per minute and 1,000 per day
account	Your account page, after confirming your email; up to 5 tokens	120 requests per minute and 20,000 per day, plus the watchlist

Account tokens look and work exactly like API keys (same headers). Limits count per key; days are UTC. Every keyed response tells you where you stand:

• X-RateLimit-Limit: requests allowed per minute for this key.
• X-RateLimit-Remaining: requests left in the current minute.
• X-RateLimit-Tier: free or account.

GET /api/pci

Returns the current Poof Chaos Index with its zone, the weighted components, the Fear & Greed reading, the combined quadrant, the daily history and the Global Poof Index.

Field	Type	Description
pci	integer	The index, 0 (calm) to 100 (maximum turbulence).
zone	object	Zone the value falls into.
zone.key	string	Stable machine-readable identifier.
zone.name	string	Display name.
zone.emoji	string	Display emoji.
components	array	The weighted inputs behind the value.
components[].key	string	Stable component identifier.
components[].label	string	Display name.
components[].weight	number	Share of the index, 0 to 1. Weights sum to 1.
components[].risk	number	Component reading, 0 (calm) to 1 (extreme). Contributes 100 × weight × risk points.
components[].detail	string	Human-readable description of the measured value.
fear_greed	object | null	Crypto Fear & Greed reading; null when unavailable.
fear_greed.value	integer	0 (extreme fear) to 100 (extreme greed).
fear_greed.classification	string	Text classification, e.g. "Greed".
quadrant	object | null	PCI combined with Fear & Greed; null when fear_greed is null.
quadrant.key	string	Stable identifier.
quadrant.emoji	string	Display emoji.
quadrant.name	string	Display name.
quadrant.text	string	Short interpretation.
history	array	One entry per recorded day, oldest first; the last entry is today's current value.
history[].date	string	UTC date, YYYY-MM-DD.
history[].pci	integer	Index value for that day.
history[].source	string	live for values recorded on the day, reconstructed for earlier days rebuilt from historical market data.
history_days	integer	Number of entries in history (up to one year plus today).
stale	boolean	true when the data missed three refreshes in a row (market data normally refreshes every 20 minutes).
updated_at	string	ISO 8601 UTC time the data was last refreshed.
methodology	string	PCI formula version, currently "v2". History only contains values of this version.
gpci	object | null	The Global Poof Index: how much of the market is in a poof state. null while it cannot be computed.
gpci.value	integer	11 (boringly solid) to 111 (total poof), on the Poof Index scale.
gpci.zone	object	key, name, emoji of the Poof Index zone it falls into.
gpci.share_in_poof	number	Share of coins in Red Flag Parade or worse, weighted by the square root of market cap (each coin capped at 15 %), 0 to 1.

Example response:

{
  "pci": 34,
  "zone": { "key": "choppy", "name": "Choppy Waters", "emoji": "🌊" },
  "components": [
    { "key": "volatility", "label": "Volatility", "weight": 0.3, "risk": 0.31, "detail": "Median daily swing ≈ 3.6%" },
    { "key": "breadth", "label": "Breadth of big moves", "weight": 0.2, "risk": 0.26, "detail": "13 of 50 coins moved more than 5%" },
    { "key": "dispersion", "label": "Dispersion", "weight": 0.15, "risk": 0.4, "detail": "Spread of 24h moves 3.1%" },
    { "key": "swing", "label": "Market cap swing", "weight": 0.15, "risk": 0.22, "detail": "Top 48 market cap +1.1% in 24h" },
    { "key": "volume", "label": "Volume surge", "weight": 0.1, "risk": 0.45, "detail": "24h volume is 2.6% of the top 48 market cap" },
    { "key": "mania", "label": "Memecoin mania", "weight": 0.1, "risk": 0.52, "detail": "Trending coins average Poof Index 52" }
  ],
  "fear_greed": { "value": 59, "classification": "Greed" },
  "quadrant": { "key": "calm-greed", "emoji": "😌", "name": "Calm Greed", "text": "Complacency mode. The quiet before the poof." },
  "history": [
    { "date": "2026-10-07", "pci": 29, "source": "reconstructed" },
    { "date": "2026-10-08", "pci": 31, "source": "live" },
    { "date": "2026-10-09", "pci": 34, "source": "live" }
  ],
  "history_days": 3,
  "stale": false,
  "updated_at": "2026-10-09T12:00:00+00:00",
  "methodology": "v2",
  "gpci": { "value": 24, "zone": { "key": "sus", "name": "Slightly Sus", "emoji": "🤨" }, "share_in_poof": 0.08 }
}

GET /api/coins

Coins with their Poof Index, filtered, sorted and paged like the Poof Index table: the same parameters, the same result. Invalid parameter values fall back to the defaults (never an error). Repeat a parameter to pass several values.

Parameter	Values	Description
tab	top | all | trending	Top 50 by market cap (default), the top 250, or trending coins.
q	text	Search in name and ticker.
zone	solid | sus | flip | red | poof	Only these zones (repeatable).
sort	poof | mcap | pc24h | pc7d | price | name	Default poof: highest Poof Index first. Coins without a value come last.
dir	asc | desc	Default desc (asc for name).
poof_min, poof_max	11–111	Poof Index range.
cap	gt10b | 1b-10b | 100m-1b | lt100m	Market cap buckets (repeatable).
ch24_min, ch24_max, ch7_min, ch7_max	number	24h and 7d price change ranges, percent.
futures	1	Only coins with a Binance perpetual.
flag	hot | longs | shorts | leverage	Positioning flags, any of (repeatable).
beta_min, beta_max	number	Market beta range.
hide_stables	1	Leave out stablecoins, wrapped tokens and pegged assets.
page	1 and up	Page number; past the last page gives the last one.
per_page	25 | 50 | 100	Page size, default 50.
limit	1–50	Older way to set the page size, used when per_page is not given. Out-of-range values are clamped.

Field	Type	Description
coins	array	Coin objects of this page, see below.
tab, sort, dir	string	The values actually used.
page, per_page	integer	The page returned and its size.
total	integer	Coins left after the filters, over all pages.
pages	integer	Number of pages (at least 1).
filters	object	The filters actually used: q, zone, cap, futures, flag, hide_stables, and the *_min / *_max ranges (null when unset).
stale	boolean	true when the list missed three refreshes in a row.
updated_at	string	ISO 8601 UTC time the data was last refreshed.
methodology	string	Poof Index version, currently "v3".
coins[].id	string	CoinGecko id, use it with /api/coins/{id}.
coins[].name, coins[].symbol	string	Display name and upper-case ticker.
coins[].price	number | null	Price in USD.
coins[].market_cap	number | null	Market cap in USD.
coins[].rank	integer | null	Market cap rank.
coins[].change_24h	number | null	Price change over 24h, percent.
coins[].poof	integer | null	Poof Index, 11 (boringly solid) to 111 (total poof); null until the coin has 31 daily closes (sorted last).
coins[].zone	object | null	key, name, emoji of the Poof Index zone; null with poof.
coins[].components	array	One entry per signal: key, label, value (the measured signal) and contribution (how far it pushes the index, in log-odds; positive = up). Empty without a value.
coins[].change_7d	number | null	Price change over 7d, percent.
coins[].has_futures	boolean	The coin has a Binance perpetual.
coins[].flags	array	Positioning flags (hot, longs, shorts, leverage).
coins[].market_beta	number | null	Beta against the top-10 market. Context, not part of the index.

curl -H "Authorization: Bearer $POOFBIT_KEY" "https://poofbit.com/api/coins?tab=all&zone=red&zone=poof&futures=1&sort=mcap&per_page=25"

Example response (one coin shown):

{
  "coins": [
    {
      "id": "bitcoin", "name": "Bitcoin", "symbol": "BTC",
      "price": 65000.0, "market_cap": 1300000000000, "rank": 1, "change_24h": 1.0,
      "poof": 14, "change_7d": 2.1, "has_futures": true, "flags": [], "market_beta": 1.0,
      "zone": { "key": "solid", "name": "Boringly Solid", "emoji": "🧱" },
      "components": [
        { "key": "vol30", "label": "30-day volatility", "value": 0.021, "contribution": -0.32 },
        { "key": "dd90", "label": "Drop from 90-day high", "value": 0.08, "contribution": -0.72 }
      ]
    }
  ],
  "tab": "all", "sort": "mcap", "dir": "desc",
  "page": 1, "per_page": 25, "total": 7, "pages": 1,
  "filters": { "q": "", "zone": ["red", "poof"], "futures": true, "...": "..." },
  "stale": false,
  "updated_at": "2026-10-09T12:00:00+00:00",
  "methodology": "v3"
}

GET /api/coins/{id}

One coin by CoinGecko id (lowercase letters, digits and dashes, e.g. bitcoin): the coin object from /api/coins plus its history.

Field	Type	Description
history	array	Daily Poof Index, up to 31 points: the last 30 days plus today, oldest first. Empty when the price history is unavailable; shorter for young coins.
history[].date	string	UTC date, YYYY-MM-DD.
history[].poof	integer	Poof Index for that day.
past.yesterday, past.last_week, past.last_month	integer | null	Poof Index 1, 7 and 30 days ago; null when unknown.
methodology	string	Poof Index version, currently "v3"; the whole history is computed with it.
positioning	object | null	Context, not scored: the coin's Binance perpetual for the last closed UTC day (day); null without a perp or when older than 3 days.
positioning.funding_rate, funding_7d	number | null	Funding as a fraction per 8 hours (the day's rates / 3) and its 7-day mean.
positioning.open_interest_usd, oi_change_28d	number | null	Open interest at the day's close, and its change over 28 days (1.0 = +100 %).
positioning.ls_accounts, ls_top_positions, taker_ratio	number | null	Long/short ratio of all accounts, of top traders' positions, and taker buy/sell volume.
positioning.liq_long10, liq_short10	number | null	Estimate: share of open interest a stated leverage model puts within 10 % below / above the price.
positioning.flags	array	Up to three plain flags, e.g. "Funding hot", "Leverage piling up".
market_beta	object | null	Context, not scored: {beta, correlation, days} of daily returns against the top-10 market over 60 days; null below 61 daily closes.
markets	object | null	Context, not scored: where the coin trades. null when nothing from the last 14 days is known; each section is empty when its data is older.
markets.native	boolean	The native coin of its own chain (no contract of its own).
markets.contracts[]	array	{chain, chain_name, address, explorer_url}: the contract on each chain (CoinGecko platform id); explorer_url is null for chains we do not map.
markets.cex[]	array	Up to 5 centralized exchanges by 24h volume, each with its biggest pair: {exchange, exchange_id, base, target, volume_usd, trade_url, trust_score}.
markets.dex[]	array	Up to 3 DEX pools by liquidity across the coin's chains: {dex, network, chain, chain_name, pair, liquidity_usd, volume_24h_usd, pool_address, url} (GeckoTerminal).
markets.updated_at	string	UTC day of the newest section. Links go to third-party sites and are not an endorsement.

An invalid or unknown id returns 404 {"error": "Unknown coin"}.

curl -H "Authorization: Bearer $POOFBIT_KEY" https://poofbit.com/api/coins/bitcoin

Watchlist

The coins on your account's watchlist. Needs an account token; anonymous keys get 403:

{ "error": "Watchlist needs an account token" }

GET /api/watchlist

{"coins", "stale", "updated_at", "methodology"}, with the same coin objects as /api/coins.

PUT /api/watchlist/{id}

DELETE /api/watchlist/{id}

Add or remove one coin by CoinGecko id. Both answer {"ok": true} and are safe to repeat. Unknown id on PUT: 404 (DELETE answers ok for any id). Adding beyond 50 coins: 409.

curl -H "Authorization: Bearer $POOFBIT_TOKEN" https://poofbit.com/api/watchlist
curl -X PUT -H "Authorization: Bearer $POOFBIT_TOKEN" https://poofbit.com/api/watchlist/bitcoin
curl -X DELETE -H "Authorization: Bearer $POOFBIT_TOKEN" https://poofbit.com/api/watchlist/bitcoin

MCP for AI agents

PoofBit is a remote Model Context Protocol server, so Claude, Cursor and other agents can read the indexes as tools. Endpoint https://poofbit.com/mcp (Streamable HTTP), same API key and rate limit as the REST API. Use the https URL in production; some clients (mcp-remote) refuse plain http unless you pass --allow-http.

Claude Code

claude mcp add --transport http poofbit https://poofbit.com/mcp --header "Authorization: Bearer YOUR_KEY"

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "poofbit": {
      "url": "https://poofbit.com/mcp",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

Claude Desktop and other stdio-only clients

Bridge with mcp-remote (needs Node.js). Add to claude_desktop_config.json; the key goes in an environment variable so the header has no space in it:

{
  "mcpServers": {
    "poofbit": {
      "command": "npx",
      "args": ["mcp-remote", "https://poofbit.com/mcp", "--header", "Authorization:${AUTH_HEADER}"],
      "env": { "AUTH_HEADER": "Bearer YOUR_KEY" }
    }
  }
}

The key goes in the Authorization: Bearer … header, as in every snippet above. The URL is /mcp (a trailing slash also works). If your client says the server needs OAuth or authentication, the API key is missing, wrong or revoked: check the header.

Tools

Tool	Input	Returns
get_chaos_index	none	The current Poof Chaos Index and Global Poof Index, same as /api/pci.
list_coins	tab, sort, limit	Coins with their Poof Index, same as /api/coins. limit defaults to 10 here.
get_poof_index	coin_id	One coin with history and context (positioning, market beta, where it trades), same as /api/coins/{id}.
explain_methodology	none	Formulas, weights, zones, quadrants, data sources and caveats.
get_watchlist	none	Your watchlist, same as /api/watchlist. Account tokens only; anonymous keys get an error result.

Try asking: “What's the Poof Chaos Index right now and which top-50 coins have the highest poof risk?”

Reading the index

Poof Chaos Index (market-wide, 0–100)

How turbulent the market has been over the last 24 hours and 7 days, 0–100, direction-neutral: a crash and a pump are equally chaotic: it describes now, not next week. Read it next to Fear & Greed as a second, independent axis. Computed over the top 50 coins by market cap, leaving out stablecoins and wrapped or staked tokens.

Zones

PCI	Zone	Color
0–20	😴 Zen Mode zen	#34a853
21–40	🌊 Choppy Waters choppy	#aef359
41–60	🎢 Rollercoaster coaster	#FFD200
61–80	🌪️ High Chaos chaos	#fb6b1e
81–100	🔥 Total Meltdown meltdown	#ea4335

Quadrants

PCI is direction-neutral, so it is combined with Fear & Greed to tell calm from panic.

Condition	Quadrant
Fear & Greed ≥ 50, PCI > 50	🤪 Euphoric Chaos euphoric Everyone's a genius and nobody's selling. Classic top-risk territory.
Fear & Greed ≥ 50, PCI ≤ 50	😌 Calm Greed calm-greed Complacency mode. The quiet before the poof.
Fear & Greed < 50, PCI > 50	🩸 Capitulation Chaos capitulation Blood in the streets. Fortunes are made (and lost) here.
Fear & Greed < 50, PCI ≤ 50	🤫 Quiet Fear quiet-fear Nobody cares. Historically, that's when patient people accumulate.

Components

Component	Weight	Measures
Volatility	30%	Median day-to-day price swings of the top 50 coins over the last 7 days.
Breadth of big moves	20%	Share of the top 50 that moved more than 5% in 24h, up or down.
Dispersion	15%	How differently the top 50 moved in 24h. Coins pulling in opposite directions = chaos.
Market cap swing	15%	How much the combined market cap of the top 50 moved in 24h.
Volume surge	10%	The top 50's 24h trading volume compared to their market cap.
Memecoin mania	10%	Average Poof Index of trending coins. Hot garbage trending = mania.

Global Poof Index (market-wide, 11–111)

How much of the market is in a poof state right now. It has no model of its own: it averages the Poof Index of the first 100 coins of the PCI universe that have one, weighted by the square root of market cap (no coin above 15%), with a power mean that lets a few coins deep in a poof pull it up. Shown on the Poof Index scale and zones. Next to it: the share of the market in Red Flag Parade or worse. It describes the present state only and gives no early warning: PCI stays the turbulence gauge (see Does it work?).

Poof Index (one coin, 11–111)

An assessment of a coin's current poof state, learned from 9 years of Binance daily prices, delisted coins included. Six signals from the coin's daily closes go through a logistic model: L (0–100) is its probability that a coin in this state is in a poof (checked against what happened to past coins in similar states). The index is 11 + L, rounded; 11 is the floor, because crypto is never 0% poof.

Signal	Measures	Higher signal
30-day volatility	How violently the price moved day to day over the last 30 days.	raises the index
Drop from 90-day high	How far the price is below its highest close of the last 90 days.	raises the index
Drop from 1-year high	How far the price is below its highest close of the last year.	raises the index
7-day move	How far the price moved over the last 7 days.	lowers the index
30-day move	How far the price moved over the last 30 days.	lowers the index
Coin age	How long the coin has had a price. Young coins go poof more often.	lowers the index

Overheating

A coin at a fresh high has no drawdown, so the six signals read it as calm, even right after a vertical run. A second reading covers that: how much further the coin ran from its 14- and 30-day lows than the typical top coin did that day. From past coins in the same spot we take the chance that the price halves within 90 days, and keep only the part above a coin that simply moved with the market. The index shows the higher of the two readings. A coin 3× ahead of the market in two weeks reads about 49, 5× or more about 65. When the whole market runs, nobody is ahead of it and the reading stays at zero.

What comes next

The six signals say how deep a coin is in a poof now. One more signal looks the other way: of past coins that stood where this coin stands, how many lost half their price within 90 days. The index is never below that share (since v3.3). It compares the coin with the other top coins that day (how volatile it is, how old, how far below its highs, how it moved this month, how closely it follows Bitcoin) and adds how far it ran ahead of the market. Volatile young coins read high, old steady ones low. Because every reading is a place among the day's coins, a market-wide move by itself changes nobody's place. The share is an average over 2021–2026 and runs from 4% to 57%.

A coin needs 31 daily closes for a value; until then lists show “—”. “Why N?” on a coin page shows how far each signal pushes the index up or down compared with a typical coin. Coin age counts from the coin's listing on Binance, as in the research; for coins Binance doesn't list, from the earlier of its all-time high and low dates. A coin with neither counts as old once we have known its price for a year.

Poof Index	Zone	Color
11–20	🧱 Boringly Solid solid	#34a853
21–35	🤨 Slightly Sus sus	#aef359
36–55	🎲 Coin Flip flip	#FFD200
56–75	🚩 Red Flag Parade red	#fb6b1e
76–111	💨 Total Poof poof	#ea4335

Context (shown, not scored)

• Size — Smaller market cap = easier to push around.
• Dilution — Fully diluted value vs market cap: how many tokens are still waiting to be dumped on you.
• Crash from ATH — How far the price has fallen from its all-time high.
• Volume anomaly — Trading volume vs market cap. Too much smells like wash trading.

Positioning (context, not scored)

Coins with a Binance perpetual show funding, open interest, long/short and taker ratios, and an estimated liquidation pressure: a model that spreads each day's new open interest over 10x / 25x / 50x leverage (50 / 30 / 20 %) at that day's price. Flags: Funding hot (7-day mean ≥ 0.05 % per 8 h), Shorts paying (≤ −0.03 %), Leverage piling up (open interest +100 % in 28 days), Crowded longs / Crowded shorts (long/short accounts ≥ 3 / ≤ 0.5), Long liquidations stacked below (half the open interest within 10 % below). The market beta is the 60-day beta against the top-10 coins. None of this feeds the Poof Index: tested, it did not make the index better.

Every payload carries methodology ("v3.3" for the Poof Index, "v2" for PCI) so you can tell formula versions apart. The Poof Index describes now, not next week.

Live track record

What we showed on a day against what the price did next, counted from the values published on that day (2026-10-10 to 2026-10-12), never recomputed. "Fell" means the price went 30% or more below that day's close within the next 30 days. It updates daily.

Poof Index zone that day	Coin-days	Fell 30%+ within 30 days
🧱 Boringly Solid	673	24%
🤨 Slightly Sus	58	24%
🎲 Coin Flip	11	36%
🚩 Red Flag Parade	8	88%
💨 Total Poof	5	100%

Does it work?

Poof Index

Tested walk-forward, year by year from 2021 to 2025: each year was scored by a model trained only on the years before it.

• What it does: it tells coins in a poof state from the rest with an AUC of 0.695. It flagged 9% of 1,221 collapses before they were obvious (before the drawdown passed 50%), a median 36 days earlier, with 1.5% false alarms. It never flagged BTC, ETH or SOL. Lead times are from the backtest, not a promise.
• Overheating: tested the same way through 2026. After an overheat reading of 36 or more, the price halved within 90 days 53% of the time (39–77% depending on the year), against 21% for all coins. So about half of these coins did not halve. On the day a collapse peaked, the index stood at 36 or more for 14% of collapses (3% before). BTC and ETH never got a reading above 23. It only sees run-ups: a slow death from a quiet top still shows up late.
• What comes next: tested the same way through 2026, coin against coin within each day. The tenth of coins this signal rated lowest halved within 90 days 10% of the time, the tenth it rated highest 33%, in that order in every year. It ranks what comes next better than the Poof Index does (AUC 0.70 against 0.65 for the index without it), which is why it is part of the index since v3.3. Before, Boringly Solid and Slightly Sus coins halved about equally often (19% and 21%); now the zones read 7%, 19%, 27%, 31% and 27% from Solid to Total Poof. Total Poof coins have mostly fallen already, so fewer of them halve again. The order holds from year to year; the level does not: in 2023 4% of all coins halved, in 2021 31%. It saw nothing unusual in LUNA or FTT before they broke.
• A heavier model: a heavier research model tested alongside (not shipped) flagged 17%, 26 days earlier, at 1.75% false alarms.
• What it can't: LUNA and FTX gave no warning: what broke them was outside price data. Hacks, frauds, unlocks and exchange trouble show up only once the price moves.
• Levels drift: the share of coins in a poof state ran from 2% to 38% a year, so the same number overstates the risk in a calm year and understates it in a crash.

Global Poof Index

It describes the present state only. Its correlation with the share of the market in a poof state at the time is 0.40 [0.11, 0.59], against 0.15 for PCI (55 monthly points). Against what happened over the following 90 days it did worse than PCI (0.12 vs 0.43): PCI stays the turbulence gauge and G-PCI is not an early-warning signal.

Poof Chaos Index

We backtested PCI on a year of daily CoinGecko data for 150 coins, fitted every constant on the first 8 months and tested once on the last 4 (June–October 2026). 95% confidence intervals in brackets.

Index	Score	Test
PCI v2	0.845	[0.634, 0.884] correlation with the same 7 days of turbulence (previous formula 0.796)
PCI v2, next week	0.42	correlation with the next 7 days: no better than assuming last week continues (0.44)

• Survivorship: the backtest only had coins that are still around.
• Calm test quarter: expect lower accuracy in a stressed market; one year of data is one market regime.
• Memecoin mania has no history to calibrate on yet; its scale will be re-anchored on live data.

Code examples

curl

curl -s -H "Authorization: Bearer $POOFBIT_KEY" https://poofbit.com/api/pci

Python

import os
import requests

key = os.environ["POOFBIT_KEY"]
headers = {"Authorization": f"Bearer {key}"}
data = requests.get("https://poofbit.com/api/pci", headers=headers, timeout=10).json()
print(data["pci"], data["zone"]["name"])

JavaScript

const res = await fetch("https://poofbit.com/api/pci", {
  headers: { Authorization: \`Bearer ${process.env.POOFBIT_KEY}\` },
});
const data = await res.json();
console.log(data.pci, data.zone.name);

Errors & caching

• 401: missing, unknown or revoked key. See Authentication.
• 403: the watchlist needs an account token.
• 404: unknown coin id (/api/coins/{id} and watchlist).
• 409: the watchlist is full (50 coins).
• 429: rate limit exceeded. Wait Retry-After seconds.
• 503 with {"error": "..."}: the data is temporarily unavailable. Retry in a minute.
• Successful responses carry Cache-Control: private, max-age=60, except /api/watchlist, which is private, no-cache. Market data refreshes every 20 minutes.
• Request bodies are limited to 64 KB.
• stale: true means the data missed three refreshes in a row; updated_at says when it was last refreshed.

Rate limits

• Free keys: 30 requests per minute and 1,000 per day. Account tokens: 120 per minute and 20,000 per day. REST and MCP count together; days are UTC.
• Over the per-minute limit you get 429 {"error": "Rate limit exceeded", "retry_after": 12} and a Retry-After header.
• Over the daily limit you get 429 {"error": "Daily limit exceeded", "retry_after": 3600}, with Retry-After counting down to 00:00 UTC.
• Data freshness is 20 minutes, so polling more than once a minute returns nothing new.
• Up to 3 new keys per network per hour.

Exchange Poof Index

A rating from 11 to 111 for centralized exchanges, built from the reserves they publish on-chain. Like the coin index, higher is worse. It reads what the wallets hold, not what the exchange says.

Signal	What it measures
Net flow, 7 days	How much of the exchange's reserves left or arrived over the last 7 days. Outflows add points, inflows take a few away.
Net flow, 30 days	The same over 30 days, to tell a one-off move from a steady drain.
Own-token share	The share of reserves held in the exchange's own token. Counted in USD; its value depends on the exchange itself.
Reserve quality	The average Poof Index of the other coins it holds, weighted by dollar value. Riskier coins add points.
Stablecoin health	The share of reserves in stablecoins trading away from their peg.
Size	Total reserves. Smaller reserves add points; larger ones take a few away.
Hacks	Past hacks, counted more when recent and half as much when the funds were returned.

Flows are counted in coins, not price: a token's move in value caused only by its price does not count as money leaving, and only tokens present on both days are compared. The exchange's own token is counted in USD only.

Flows are priced at the lower of a token's two prices, so a token whose amount was re-based does not read as money leaving.

A sudden jump in the 7-day outflow is scored at the day before's milder reading until it holds a second day: one-day jumps are as often an exchange moving its own wallets as a run. On past data this removed 37% of red episodes and left the readings after known hacks and runs the same or a day later.

Does it work?

• What it shows: how fast coins are leaving the wallets an exchange has published, and how fragile those visible reserves look. After the Bybit hack of February 2025 it went from 18 to 62 in two days.
• What it can't: it reacts, it does not warn. A week before seven known stress events the exchange in question sat among the calmest. A low number means no visible run today, not a safe exchange. It is not a probability of failure.
• What nobody outside can see: what an exchange owes its customers, its off-chain and bank holdings, wallets it has not published, loans to related companies. FTX failed on exactly those.
• Not tested on real failures: public reserve history starts in November 2022 and holds none of the exchanges that failed, so the weights are hand-picked, not fitted. On the 63 exchanges we could test, most of what the index ranks is size: small exchanges lose their visible reserves more often.
• A high number is not always a run: an exchange moving coins between its own wallets, or changing its published wallet list, looks the same from outside.

Reserves unknown means we cannot rate the exchange: it has not published its wallets, we have less than 30 days of history for it, or its data is not fresh.