Real-time trading notifications
powered by on-chain intelligence

A single webhook endpoint that accepts our POST payloads (JSON, bearer-token auth). Broadcast notifications require zero matching logic — just fan out to all users. For personalized notifications at scale, you need a batch endpoint that accepts up to 1,000 recipients per call, or the ability to match a segment signal against your own user-token holdings internally.

Broadcast: ~10–15 per day (one POST per signal, you fan out internally). Personalized for small tokens: batched — 1 POST per 1,000 recipients (a token with 5K holders = 5 calls). For mega-popular tokens (>50K holders): a single segment signal — we send the filter criteria, you match internally. Total: typically under 100 webhook calls per day in steady state.

Each signal is a JSON POST with: signal_id, signal_type (convergence, whale_move, etc.), priority (P0–P2), chain, token details (address, symbol, price, change %), headline (pre-built notification copy), trader data (rank, PnL, win rate, trade size), summary stats, a CTA object with swap action + token address + chain ready for deep linking, wallet_id, and timestamp. Personalized signals include an additional context object.

Under 60 seconds end-to-end. On-chain trade hits the stream, FOMO detector processes it in a 5-second batch window, checks rank + convergence state (single-digit ms DynamoDB lookups), and the delivery Lambda POSTs to your webhook. Personalized adds a few seconds for fanout. Segment signals are equally fast since it's a single POST.

Nothing for Phase 1 (broadcast) — we derive everything from on-chain data. For personalized: a mapping of which wallet addresses belong to your users so we can target by wallet_id. Alternatively, with the segment model you just need to answer which of your users hold a given token internally. No PII, no user accounts, no API keys to your systems — it's a one-way push.

Signals queue in our internal stream with 24-hour retention — nothing is lost. Delivery retries with exponential backoff (1s → 2s → 4s). Failed deliveries go to a dead-letter queue for manual replay. We can replay from the candidates stream during prolonged outages.

Three layers: (1) Global throttle — max ~15 broadcast signals per day, configurable per type. (2) Dedup — same token + same signal type won't re-fire within 2 hours unless it escalates. (3) Per-wallet budget — max 5 personalized notifications per day per wallet. All thresholds are tunable based on your CTR data.

Whatever the leaderboard pipeline supports — currently derived from our Flink pipeline processing multi-chain DEX trades. Each notification includes a chain field so you can filter or route by chain on your end.

Yes. Each payload has a signal_type and priority field. Filter by type (only convergence, only whale moves), filter by priority (only P0/P1), or apply your own logic before surfacing to users. We deliver the signal — you decide what to show.

Our cost scales with signals, not users. Broadcast: 1 POST regardless of user count — you fan out internally. Personalized: tiered strategy. Small tokens get individual targeting, mega-popular tokens get a segment signal (one POST, you match internally). We only index wallets that actually trade on-chain (~1–2% of your user base). The other 98% get broadcast only at zero incremental cost.

Bearer token in the webhook header. We store your API key in AWS Secrets Manager, you validate the token on your endpoint. Standard, simple, nothing proprietary.

Signals come from proven profitable traders ranked by real PnL on a public leaderboard. This is on-chain execution data from wallets that have actually made money — not sentiment analysis or social hype. Convergence (multiple independent profitable wallets buying the same token) is the strongest signal because it's hard to fake.