Your finance agent is confident and unaccountable. Fahali is the early-warning read it checks its own conviction against: a judged risk read with its confidence, the inputs it is missing, and a signed receipt. Read-only. Observation, not advice.
See it before you sign up. The scored track record is a public endpoint:
curl https://app.fahaliai.com/api/track-record/lead-time
Returns, per engine, how many hours ahead it tended to warn on its correct calls, with the misses and the sample size. No key, no signup. That is the one thing a price feed cannot give an agent.
Connect MCP-compatible agents to market, risk, and evidence tools through streamable HTTP or SSE. The point is not more prices; it is a market read your agent can challenge.
MCP: mcp.fahaliai.com/mcp
MCP server docs →Use the OpenAPI contract when your application needs live market context outside an MCP client. Public aggregates are available before sign-in; deeper reads use your Fahali key.
OpenAPI spec →Fahali keeps missing inputs visible and attaches provenance information to issued reads. It is market observation, not execution or financial advice.
llms.txt →Terminals were built for human eyes. Generic search gives your agent unaccountable text. Fahali gives it claims it can verify.
Fahali publishes, per engine, the median lead time on its correct material warnings: how many hours ahead of the move it fired, scored against what happened, misses included. Your agent can weight a signal by its engine's proven lead, not just the call. Free on the developer key.
See the lead-time record →Every verdict response carries a SHA-256 receipt and a provenance root. When your agent quotes Fahali, it can attach cryptographic proof of exactly what was said, when.
Every predictive signal is judged against a disclosed, versioned methodology at its own horizon, and every judged outcome has a public replay URL your agent can cite. Analytical engines are labelled observation-only rather than graded as forecasts. Wrong calls stay on the record.
See the live tape →The 72h forecast records its full probability distribution and is Brier-scored against realized outcomes. Your agent can weight our probabilities by our measured calibration, not our marketing.
Missing data says so. Quiet markets resolve as unresolved, never wins. No-signal states are explicit. Your agent never has to guess whether silence means calm or failure.
Read the methodology →A free developer key gives you a small, practical evaluation path. Use it to test whether Fahali's market-read structure improves your agent before you commit to a deeper workflow.
Sign in at app.fahaliai.com and open the Developer screen to generate a free key. The dashboard shows the current connection details for your client.
Open the Developer screen →Prefer email? Write to robert@futurelegendsai.com. Existing subscribers already have API access at Professional tier and above; the same screen shows your key.
Three commands. No SDK, no docs detour. If step 3 returns a verdict with a receipt, you know exactly what Fahali is.
# 1 — no auth: is the platform live and covering the market right now? curl https://app.fahaliai.com/api/public/stats # 2 — no auth: what moved in the last 24h? curl https://app.fahaliai.com/api/market/movers # 3 — your free key (Developer screen): a judged market verdict, with receipt curl -H "Authorization: Bearer YOUR_KEY" \ "https://app.fahaliai.com/agent/verdict?symbol=BTCUSDT"
Step 3's response carries the read and the receipt: missing inputs named, metric/window fields explicit, a SHA-256 provenance hash. Pipe it to your agent and let it challenge the reasoning. That is the intended use.
Connecting an MCP client instead? The Developer screen generates the exact config for your client (Claude, Cursor, or any MCP-compatible agent): one paste, ~25 read-only tools.
Fahali supports streamable HTTP and legacy SSE. Choose the client in your Developer screen, then use the generated configuration rather than copying a stale snippet from a marketing page.
Streamable HTTP: https://mcp.fahaliai.com/mcp SSE: https://mcp.fahaliai.com/sse Auth: Fahali key or the supported connect flow for your client
MCP and the read API start free. Depth and volume scale with the lane. Human app tiers are separate and unchanged.
50 calls/day. Verdicts, the tape, replays, the judged track record, and the verified lead-time record. Signed receipts included. No card.
Start now →10,000 calls/mo, then $5 per 1,000. Core intelligence tools: verdicts, 72h forecasts, alerts, market snapshots.
100,000 calls/mo, then $3 per 1,000. Every read tool: capital flow, dark-pool proxy, contagion, smart-money reads, shock tests, plus SSE streaming.
1,000,000 calls/mo, then $2 per 1,000. Webhooks, schema-change notice, dedicated support channel.
Activating a paid lane: self-serve checkout is rolling out. Until it lands, email robert@futurelegendsai.com and we activate same day. Enterprise (dedicated MCP instance, SLA, custom ingress) remains a conversation.