fahali_get_market_verdictLABSCO SUMMARY
The misses stay on the record, which is what makes the hits worth quoting.
Predictive signals are graded at their own horizon against a published method, and analytical engines are labelled observation-only rather than scored as forecasts. That distinction, plus a lead-time record you can read before signing up, is the difference between a confidence number and a track record.
What it is
A read-only financial-risk layer for agents. Your agent already has prices; this says what they mean for capital at risk — portfolio risk, stress losses, crypto risk, contagion, crash precursors and per-symbol verdicts — and returns each read with its confidence, the inputs it lacked, and a receipt. It observes and explains; it never says what to buy or sell, and there is no order routing anywhere in it.
What you get
- A per-symbol verdict with direction, confidence, reasoning, expected move and horizon
- The outcome-scored lead-time record — per engine, the median hours ahead of the move on correct material warnings, with the misses kept in the same record and readable without a key
- Probabilistic forecasts over each signal's own registered horizon of 6 to 72 hours, with an uncertainty cone and a Brier score behind them
- Flash-crash precursor strength, or an explicit abstention when the data is not there
- The judged scorecard with per-horizon base rates and lift, replayable against a disclosed, versioned method
- Cross-asset tail dependence and the clusters that move together, for contagion rather than single-name risk
- Absorption inferred from public microstructure only, and an order-flow imbalance proxy with no invented institutional-versus-retail split
- The current market regime, and a portfolio stress-tested against a scenario you name
- A SHA-256 receipt and provenance root on every verdict, so a quote can carry proof of what was said and when
- Absence stated as absence: quiet markets resolve as unresolved rather than as wins, and whale tracking says there is no on-chain source instead of inventing positions
Requirements
Nothing to install for the hosted path: streamable HTTP at https://mcp.fahaliai.com/mcp, with the legacy SSE endpoint still served. Authenticate with OAuth 2.1 or a static sk_live_ bearer key; a free developer key allows 50 calls a day with no card. The repository also runs as a local stdio proxy that forwards each call upstream with FAHALI_API_KEY set, and the tier is enforced on the server side either way. Coverage is crypto around the clock and US equities and ETFs during market hours.
Setup effort
Add the hosted endpoint — paste it into your client
USE CASES
01
Checking risk before a position is opened
Ask whether a crash precursor is firing, read the verdict with what data is missing, and weigh it by how early that engine usually warns.
02
Deciding how much to trust a signal
Read the judged scorecard with its per-horizon base rates and lift, then the forecast with its uncertainty cone rather than a single number.
03
Seeing what would go down together
Pull the cross-asset tail dependence and the correlated clusters, so a hedge is not sitting inside the same cluster as the position.
04
Putting a number on a bad day
Stress the portfolio against the scenario you are worried about and read the loss rather than guessing at it.
05
Reading the tape without pretending to see more than you do
Take absorption from public microstructure and the order-flow imbalance proxy, both explicit about what they cannot observe.
10 tools
Listed in project docs. Your client may show a different list.
fahali_get_lead_timefahali_get_forecastfahali_get_flash_crash_riskfahali_get_track_record_scorecardfahali_get_contagion_mapfahali_get_microstructure_proxyfahali_get_capital_flowfahali_get_market_regimefahali_run_shock_testBEFORE YOU INSTALL
What to know first
Tools you get
10 tools named in the project's documentation.
Requires
A Fahali account: approve the OAuth sign-in ("Connect with Fahali") when the client connects, or send a free developer key from app.fahaliai.com/developer as a Bearer token in the Authorization header
Where it runs
All 3 documented clients connect to a hosted endpoint.
INSTALLATION
Connect your client
Server setup
Claude Desktop, ChatGPT
ENDPOINT URL
https://mcp.fahaliai.com/mcpAdd this hosted endpoint to your MCP client.
MCP Clients
Cursor
CURSOR · JSON
{
"mcpServers": {
"fahali": {
"url": "https://mcp.fahaliai.com/mcp"
}
}
}
