gamma_profile on Hyperliquid GEX

What It Is

The live board is https://cryptodataapi.com/quant-gamma.

The search is gamma_profile. The page prints that field in the “what this is” block. The MCP tool is get_gamma_exposure. The REST call is GET /api/v1/quant/gex?symbol=BTC.

I am Sam Deering. I built this so an agent can read where dealer forced-unwind fuel stacks by price. Agents read. They do not trade. We are not a broker.

gamma_profile is the liquidation-density histogram of classified Hyperliquid market-maker inventory by price.

The page calls it the gamma profile. Same idea. Same field.

Bins are 0.5 percent of mark. The window is plus or minus 50 percent of mark. Tall bars are clusters where forced unwinds stack up. Those are levels price gravitates to and accelerates through.

The page says this is the perpetuals analog of options GEX. Perps have no literal options gamma. This is a behavioral analog. It is built from the exact on-chain liquidation price of every classified dealer account at or above $100k.

The API docs for GET /api/v1/quant/gex name the same field. They say gamma_profile is MM liquidation-density by price — where forced unwinds cluster. llms.txt repeats that wording.

The profile answers one question. If price reaches level L, how much notional is force-unwound there? It does not answer the probability that price reaches L. Options markets extract that second answer from strikes. Perps have no strikes. There is no implied-distribution shortcut here.

Where to See It

Open https://cryptodataapi.com/quant-gamma. Look for the MM gamma profile / liquidation density card on each coin.

The human page is a free launch preview. The page says the preview is about 30 minutes delayed. The API is real-time.

The REST path is Pro:

curl -H "X-API-Key: cdk_live_your_key" \
  -H "User-Agent: cryptodataapi-python/1.0" \
  "https://cryptodataapi.com/api/v1/quant/gex?symbol=BTC"

The MCP tool that maps to that path is get_gamma_exposure. That name is on https://cryptodataapi.com/llms.txt and on https://cryptodataapi.com/ai-agents/mcp-server. The gamma page itself does not print the tool name.

The full gamma profile is archived per snapshot as gamma_exposure. The page dates that archive to 6 Jul 2026 (2026-07-06). The archive curl on the page is:

curl -H "X-API-Key: cdk_live_your_key" \
  -H "User-Agent: cryptodataapi-python/1.0" \
  "https://cryptodataapi.com/api/v1/backtesting/snapshots?data_type=gamma_exposure&start=2026-07-06&limit=1000"

That archive is Pro Plus. The live board is a 3-coin teaser. Free preview coins are BTC, ETH, and SOL. The API universe is 176 coins.

How to Read It

Read the histogram as density by price. Tall bars mean more dealer notional force-unwinds if mark reaches that bin.

The page marks some density prices as upside unwind and some as downside unwind. Read those labels with the bars. Do not invent a probability from a tall bar.

Then read companion fields as separate answers:

gamma_flip — one crossover price. Long-liq below crosses short-liq above. Docs cap it to ±30% of mark. No in-band crossover means null.
Dealer net inventory — the page label for mm_net_delta. Aggregate long minus short across classified dealers in that coin.
Regime flag — amplify, dampen, or transitional. Read the flag before you treat a tall bar as a trade idea.
distribution_context — trailing-30-day ranks of density and related series. A raw density number means little on its own. The percentile lives in that block, not inside the histogram bars.

The profile is the density stacked around prices. The flip is one price. Do not collapse them.

What It Is Not

This is not gex_veto. gex_veto is the amplify stand-down. When the regime flag reads amplify, agents stand down fresh longs until the flag clears. That rule is a sibling post, not this one — see gex_veto gamma flip Hyperliquid. gex_veto is not a JSON key. Do not invent a boolean.

This is not a distribution_context lead. That block ranks today's near-mark fuel, flip distance, and MM skew against the coin's own 30-day history. That block is a sibling post, not this one — see distribution_context on Hyperliquid GEX. Read that block there. Do not rewrite it here.

This is not the HMM market-regime object on /quant/market. That engine is a different head. Do not collapse histogram bins into HMM posteriors.

This is not candles. Bars cannot tell you where dealer forced-unwind density stacks.

This is not an options greek. The page says it is a behavioral analog.

This is not a probability that price reaches a level. It is a distribution of forced flow by price.

This is not gamma_flip by itself. The flip is listed as its own field.

This is not liquidation_map. That dataset is the full-universe liquidation-density heatmap the gamma regime is read against, archived on its own schedule.

This is not hl_trader_positions. That dataset is the classified account universe the gamma is derived from.

What the Board Showed

Fetched from https://cryptodataapi.com/quant-gamma on 10 Sep 2026, afternoon Brisbane time (AEST, UTC+10). The page says this preview is about 30 minutes delayed. The API is real-time. These figures are the page. A later call will differ.

Header on this fetch: 3,460 accounts tracked. 222 classified dealers. Free preview BTC · ETH · SOL. 176-coin universe in the API.

Coin	Mark	MM net inventory	Gamma flip	Dealers	MM long / short
BTC	$78,304.50	SHORT $78,776,172	$93,593.11 (+19.52%)	98	$19,421,997 / $98,198,169
ETH	$2,474.75	SHORT $134,372,079	$2,377.93 (−3.91%)	94	$64,830,728 / $199,202,808
SOL	$101.78	SHORT $38,084,264	$73.27 (−28.02%)	80	$10,974,414 / $49,058,679

Printed density prices on the profile cards this session — the histogram's visible bins on the HTML card, labeled upside unwind or downside unwind, not a full ±50% dump of every 0.5% bin:

BTC: $92,987, $88,288, $86,722, $77,326, $76,151, $74,977, $72,627, $70,670, $68,712, $64,797, $64,014.
ETH: $2,951, $2,407, $2,382, $2,345, $2,233, $2,221, $2,097, $2,085, $2,073, $2,060, $1,998.
SOL: $117.31, $105.60, $98.99, $98.48, $94.91, $93.39, $86.77, $86.26, $83.72.

The page also printed trailing-30-day ranks under distribution_context (orientation only — not this post's lead): BTC fuel density P74, flip distance P50, MM skew P36, 30.0d. ETH P94 / P5 / P94, 30.0d. SOL P75 / P94 / P55, 30.0d. Read those ranks on the sibling piece. Read the bars here.

One Concrete Live Example

Call GET /api/v1/quant/gex?symbol=BTC.

Read the BTC coin's gamma_profile.

Confirm mark. On this page fetch it was $78,304.50.
Read the density bins. On the HTML card the visible stack included $92,987 above mark and $77,326 / $76,151 / $74,977 below it.
Note upside-unwind versus downside-unwind labels on those bars.
Read gamma_flip as a separate field. On this fetch it was $93,593.11 (+19.52%).
Read mm_net_delta / net inventory. On this fetch SHORT $78,776,172 across 98 dealers.
Read the regime flag before you treat a tall bar as actionable.
If you need history context, open distribution_context. Do not invent percentiles inside the histogram.

I called the REST path in this session with a named User-Agent and no key. HTTP 401. Body error: api_key_required. Message pointed at POST /api/v1/auth/keys. I did not invent a payload. I did not copy bins out of the 401.

A second call with an invalid X-API-Key also returned HTTP 401. Body error: invalid_api_key. That is still not a Pro payload.

For Agents: MCP Tool Order

Call get_gamma_exposure when the user asks for gamma_profile.

Do not open with a candle poll. Bars cannot tell you where dealer forced-unwind density sits.

Do not open get_order_book for this query. The book is a different page.

Tool order:

list_capabilities — no key needed.
create_free_api_key — only if you lack a key. Use a real email.
Confirm that email. Confirming lifts the same key from 100 to 1,000 requests a day and switches Pro on for 24 hours.
get_gamma_exposure — Pro. Maps to GET /api/v1/quant/gex.
Read gamma_profile on the coin. Report the density by price.
Report gamma_flip and mm_net_delta as separate fields.
Read the regime flag. If it is amplify, the sibling gex_veto stand-down applies. Do not invent a gex_veto JSON key.
Read distribution_context only when you need trailing ranks. That is not the histogram.

A 403 on that tool is the expected free-tier gate. Read the body. Do not retry it. The docs say a JSON 403 carries required_tier and pricing_url. A different 403 is HTML from the CDN. Python's built-in urllib is refused before the API sees the call. Set a User-Agent that names your client.

Backtesting snapshots are not on the MCP tool list. Use REST for GET /api/v1/backtesting/snapshots?data_type=gamma_exposure. That path is Pro Plus. The page says the archive keeps the full gamma profile at 5-minute resolution.

REST

curl -H "X-API-Key: cdk_live_your_key" \
  -H "User-Agent: cryptodataapi-python/1.0" \
  "https://cryptodataapi.com/api/v1/quant/gex?symbol=BTC"

Omit symbol for the full universe on Pro. The page says 176 coins in the API. The free HTML preview is three coins.

Pro unlocks /quant/gex and /quant/positioning for the Hyperliquid universe. /quant/positioning splits notional by trader type — market_maker, whale, other, and all — with smart_money and high_leverage as separate orthogonal overlays, not a breakdown of any trader-type bucket. The tagged liquidation-map archive is Pro Plus.

Archive:

curl -H "X-API-Key: cdk_live_your_key" \
  -H "User-Agent: cryptodataapi-python/1.0" \
  "https://cryptodataapi.com/api/v1/backtesting/snapshots?data_type=gamma_exposure&start=2026-07-06&limit=1000"

Limits the page states: full-universe liquidation_map archives every 5 minutes. gamma_exposure snapshots — including gamma_profile — start 2026-07-06. distribution_context specifically was added later, on 2026-07-10. GET /api/v1/backtesting/snapshots/types is the authoritative live source for exact per-type coverage. None of that history is backfillable. Positioning is point in time. The dataset is forward-only and grows every week. Percentiles stay null while a coin's history is warming (under 2 days). A study of forward return, realized vol, and tails conditioned on the regime flag is still in progress. Findings land when the forward-only archive is deep enough.