of warning before a 11.4% fall became visible. The price peaked at 19:00 UTC; the warning was on the record before the fall could be seen.
Hours of warning. Every miss on the record.
Type what you hold. See what is changing around it, how serious it is and how much evidence stands behind the call, then keep it under watch after you close the page.
- Free to read, no sign-up
- Read-only, never touches capital
- For people and AI agents
- crypto instruments scanned every cycle
- 573
As of 11 Oct 2026 09:17 UTC, measured by the scan scheduler: 573 of 573 liquid crypto instruments scheduled, 486 with live market data in the last cycle. Source (JSON)
Every risk vendor says it warns you early. Ask them for their misses.
A 46.2% fall from the 15:00 UTC peak. Warnings arrived only once the fall was already visible. That is detection, not warning. It counts against us, and it stays on the record.
Last 30 days, counted per trading pair: 133 large falls. 78 had no usable early warning, 8 were warned only marginally, 47 were warned early.
These count falls, not warnings: how often a warning is followed by a fall is in the record below. A single case is one fall, not a typical lead time; the median is in the record. Case files as computed 11 Oct 2026. Definition: a major event is a fall of >= max(8%, 3x the coin's 24h volatility) within 24h of a 72h high; a catch requires an EARLY, ADVERSE, IN-HORIZON, UNUSUAL-for-that-engine warning before the fall was visible. Every case file, misses included (JSON) →
Reading the market.
What is changing, how serious it is and what is connected, on the market as it is right now. When the evidence is too thin to make a call, the read says so instead of guessing.
Type what you hold at the top of the page and this read re-runs on it, with the market-wide break and contagion reads alongside.
You name what you hold. You hear when risk changes around it.
- Hear it earlyA warning when risk starts building around a position, and a public record of how often that warning came early and how often it did not.
- See what's connectedWhich of your positions sit in the path when stress in one market starts pulling on another.
- Only what changedThe situation stays open. You hear when it escalates, eases or changes shape, not every time a number ticks.
Every holding is read along the same six questions.
- ConcentrationAre several holdings quietly becoming the same risk?Positions that move together under stress, read from how they actually co-move rather than what sector they are filed under.
- ContagionIs stress somewhere else starting to pull on what you hold?Which markets are falling together on the downside, and whether your positions sit in that path.
- LiquidityIs it getting harder to get out?Trading conditions thinning around an asset, estimated from market data and labelled as an estimate.
- FlowIs money leaving faster than usual?Unusual movement of money into and out of sectors and assets, against what is normal for each.
- RegimeHas the market changed character?The current market state and how fresh that read is, so a calm-market assumption is not carried into a stressed one.
- ShockWhat would a past stress move do to this book?Your positions run through historical stress moves, as a scenario, not a forecast.
From the first sign to the record.
- DetectNotice risk building around an asset or a market.
- ConnectLink it to what else is moving and to what you hold.
- JudgeSay how serious it is and how much evidence stands behind it.
- WatchKeep the situation open and report only when it changes.
- VerifyScore the call against what prices did, beside the base rate.
- RememberKeep the outcome on the record, hits and misses alike.
How much warning, measured. The misses included.
Every warning is registered before the outcome exists, then judged against realized prices and against how often the same fall happens anyway. Precision is published beside recall.
A call, and what happened next.
One open call with its claim and horizon, and one judged call with what the market did next. The judged call stays on the record whether it was right or wrong.
Give your agent a market-risk check it can cite.
Before an agent allocates, rebalances or writes a thesis, one call returns what is changing, the evidence, and what is still uncertain. Thin evidence comes back labelled as thin.
One call
# A market-risk read. No key needed to try it. # Add ?symbols=<your assets>, or omit it for today's measured book. curl "https://app.fahaliai.com/api/market-judgment" # Or connect any MCP client https://mcp.fahaliai.com
- Works with Claude, ChatGPT and any MCP client.
- Read-only: no tool can place an order or move capital.
- Every agent verdict carries a SHA-256 integrity hash.
- 50 calls/day - verdict, tape, replays, track record
- MCP + REST, SHA-256-hashed verdict receipts
- Public methodology + judged record
- 10,000 units/mo included, no hard cap — most calls cost 1 unit
- Core intelligence tools - verdicts, 72h forecasts, alerts, snapshots
- Email support
- 100,000 units/mo included, no hard cap — same per-call prices as Starter, ten times the allowance
- ALL read tools - capital flow, microstructure proxy, correlation chains, unusual money movement, shock tests
- SSE streaming + priority support
- 1,000,000 units/mo included, no hard cap
- Everything in Pro + webhooks + schema-change SLA
- Dedicated support channel
Read for free. Pay when you want it to keep watching.
For anyone checking the market
$0See the live market.
The live read, forecast, break layer, contagion and the public record. No card, no account.
See the marketFor the investor with positions to keep an eye on
$19/monthKeep your assets under continuous watch.
State-change alerts on the assets you hold, saved cases and the full instrument read.
Watch my assetsFor the active professional
$79/monthGo deeper on what is changing.
Deeper evidence, cross-asset intelligence and read-only API access. 3-day trial.
Start ProfessionalFor the portfolio manager
$299/monthRun the whole portfolio through Fahali.
Portfolio-level shock tests, every surface of the app and the decision trail.
Start EliteFor the risk desk
$1,499/monthOne picture of risk for the whole desk.
Shared cases, alerts and an audit trail where the risk team already works.
Open the desk workspaceFor teams building agents
from $999/monthMarket-risk checks inside your own systems.
1,000,000 units/mo included, no hard cap. Everything in Pro + webhooks + schema-change SLA
Build with FahaliEnterprise · unlimited seats, MCP integration and a custom data agreement, from $4,999/month. Talk to us →
- Registered firstEach warning is written down with its claim and horizon before the outcome exists.
- Judged on pricesScored against what the market did, on a volatility-scaled bar.
- Against the base rateRiding a falling market earns nothing.
- Read-only by designNo order routing, no path to capital.
Frequently asked questions
What is Fahali?
Fahali gives you hours of warning before market risk reaches your assets, and publishes every miss beside every catch. It watches crypto markets continuously and US equities on a rotating schedule during market hours, from Future Legends Inc. Every warning is registered with its claim and horizon before the outcome exists, then scored against realized prices, so the record cannot be edited after the fact. Investors put a portfolio under watch, desks share one picture of risk, and AI agents call it as a market-risk check before they act. It is read-only by design: no orders, no path to your capital.
How much warning does Fahali give before a market drawdown?
A median 3.93h of warning before 405 of 1,710 qualifying crypto drawdown events (recall 23.68%), with precision 8.36% against a 5.07% base rate. It also warned in 13.1% of sampled quiet 8h windows, where no qualifying event came within 8h either side. That is one published stratum (crypto, 8h horizon), not a figure for every market. Record as measured 11 Oct 2026. The current measurement is always served at https://app.fahaliai.com/api/track-record/lead-time.
How are Fahali's warnings judged?
Each warning is written to the record with its direction or event, its horizon and its confidence before the outcome exists, so it cannot be edited or backfilled. When the horizon closes it is judged against realized prices on a volatility-scaled bar, compared with the base rate of the same event happening anyway, and given a measured grade from beats chance to worse than chance. Precision is always published beside recall.
Can AI agents use Fahali?
Yes. Agents connect through the Model Context Protocol at https://mcp.fahaliai.com or through the REST API, and every agent verdict returns with a SHA-256 integrity hash, so it can be checked later. The free agent key includes: 50 calls/day - verdict, tape, replays, track record; MCP + REST, SHA-256-hashed verdict receipts; Public methodology + judged record. Paid agent plans: Agent Starter $49/month, Agent Pro $199/month, Agent Scale $999/month.
How much does Fahali cost?
The public market read on this page is $0, with no card and no account. Plans for people: Retail $19 per month, Professional $79 per month, Elite $299 per month, Desk $1,499 per month, Enterprise $4,999 per month and up. Professional starts with a 3-day trial. Current prices are always served at https://app.fahaliai.com/api/tiers.
Does Fahali trade or give investment advice?
No. Fahali is read-only by design: it has no order routing and no path to capital. It publishes observations about market risk and the measured record behind them, not recommendations to buy or sell.
Can I verify Fahali's record myself?
Yes, and that is the point. The method is public and the record is served as JSON per stratum, with sample size, base rate, recall and a measured skill grade. Every call stays on it, hits and misses alike. Each agent verdict carries a SHA-256 integrity hash, so any quoted call can be checked against the record by your own team.
Start with what you hold.
The live read is free, with no sign-up. Put it under watch when you want it to keep looking, or give your agent a free key.