Snipe Pump.fun tokens minutes before they migrate
Every Pump.fun token graduates at ~85 SOL of bonding-curve reserves. Subglow gates your copy-trade rules on that curve state — so you only fire when a tracked wallet buys a token that's mathematically close to migration. No chasing graduated AMM candles. No dusty 73%-complete rugs. Just the final sprint.
Three checks, one trade
Migration sniping on Subglow is not a separate bot — it's a gate on top of our existing copy-trade engine. Every migration buy clears three independent filters before it executes.
Your selected wallet — from Rising Stars, the leaderboards, or Discover — fires a buy on a Pump.fun mint. Our Rust monitor sees the swap via RPC or gRPC within ~200ms.
We pull the bonding-curve PDA state in one round-trip (cached 10s), compute completion from real_sol_reserves ÷ graduation threshold, and compare against your configured minimum. If the curve is cold, the trade is skipped.
If both gates pass, we build the pump.fun buy instruction ourselves — compute-budget tuned, optional Jito bundle, ATA pre-warmed. No Jupiter quote detour. Typical land time: 150-300ms end-to-end.
The 85-SOL threshold is a hard edge
Pump.fun's bonding curve is a constant-product AMM with virtual reserves. The math doesn't care about vibes — the moment real_sol_reserves reaches the graduation constant, the program sets the complete flag and the LP migrates to PumpSwap. Every token that ships crosses this exact line.
Pre-migration means buying while the curve still has capacity, at bonding-curve price, seconds before liquidity unlocks. The moment graduation fires, the first PumpSwap candle typically re-rates 50-200% because the pool's effective price is set by bonding-curve virtual reserves, not actual AMM depth.
Post-migration means chasing the PumpSwap candle. You're now competing with every other sniper bot that saw the migration instruction, your slippage is higher, and the early entrants already own the float. Phase 2 will add a post-migration chase mode as an option — but pre-migration is the cleaner play in 99% of cases.
The 85% default threshold is deliberately not 100%. Setting it too close to graduation shrinks your entry window to seconds. 85% gives you minutes of exposure before the pump, a much better risk/reward than trying to be the absolute last buyer on the curve.
The pre-migration rotator
Here's the config we recommend for a first migration rule. Every value is one click in the configure UI.
- StrategyFixed · 0.3 SOL
- Migration snipeON · ≥ 85%
- Min trade size0.1 SOL
- Anti-rugModerate
- New position onlyON
- Take profit+60%
- Stop loss−25%
- Trailing stop15% off peak
- Copy sellsOFF
- SlippageAuto
A tracked wallet buys a Pump.fun token. Our engine checks the curve — 87% complete. Gate passes. Anti-rug clears. We fire a 0.3 SOL buy through the direct Pump.fun builder. You hold through migration. Token re-rates 80% on the first PumpSwap candle — take-profit triggers and exits. On the other tokens that stalled at 60%, the gate kept you out entirely. That filter is doing most of the work.
From pre-migration to graduation-moment sniping
Subglow's migration sniper is shipping in phases. We build the valuable-and-safe layer first, then add latency and expressiveness on top. Every phase composes with the previous one — your Phase 1 rules keep working as Phase 2 features ship.
Bonding-curve tracker (10s TTL cache), migration-only filter gate on CopyConfig, direct Pump.fun builder execution path. Today. Configurable via the Smart Money Alpha section of the rule editor.
Yellowstone gRPC program-subscription on the Pump.fun `withdraw`/migrate instruction so we see graduation in the same slot, plus a native PumpSwap pool builder for the first-candle buy. Multi-tx Jito bundles for atomic approve+swap.
For hot migration candidates (≥ 95% complete with whale volume), we encode + sign a buy transaction in advance so we can broadcast in the same slot the migration trigger fires. Target: sub-50ms detection-to-land.
Optional strategy for tokens that survive the first 3 minutes on PumpSwap with ≥ N tracked wallets still buying. Different TP/SL profile. Opt-in per-rule — pre-migration stays the default.
Before you enable it
What is 'migration sniping' on Pump.fun?+
Every Pump.fun token lives on a bonding curve until it fills to ~85 SOL of real reserves, at which point it 'graduates' — migrates to the PumpSwap AMM where price action tends to re-rate violently as the LP becomes liquid for larger flows. Migration sniping means buying a token in the last stretch of its bonding curve (~85% complete) so you're positioned for that re-rate, rather than chasing it after migration has already happened.
How does Subglow's migration sniper actually decide to fire?+
Instead of scanning every Pump.fun mint (which would burn RPC credits and catch rugs), we gate our existing copy-trade signal on the curve state. When a wallet you track buys a Pump.fun token, our Rust engine asks the bonding-curve tracker: is this mint ≥ X% complete? If yes, it fires the buy; if no, it skips. So every migration snipe has two independent confirmations: (1) a top trader you selected just bought it, and (2) the token is mathematically close to graduation. The default threshold is 85%.
How fast is execution?+
The buy path uses our direct Pump.fun builder — we emit the global:buy instruction ourselves, with compute-budget tuning and Jito tip/bundle support, bypassing the Jupiter quote+swap round-trip (~250-500ms saved). The gate itself is cached in-memory (≥10s TTL), so cache-hits add <1ms to the decision path. End-to-end from swap detection to transaction broadcast is typically 150-300ms on Helius RPC; on gRPC (coming in phase 2) we target sub-100ms.
Why gate on a top trader's buy instead of just scanning every mint?+
Sniping ANY mint at 85% complete is a losing strategy — most of them rug, most of them never actually graduate, and the ones that do are already tip-heavy contested by other bots. Gating on a top trader's buy adds information: you're only sniping tokens where an on-chain wallet with a proven win rate just decided to deploy capital. That's what turns 'migration sniping' from a lottery into a strategy you can actually backtest and tune.
What's the recommended TP / SL pairing?+
Pre-migration plays are momentum trades, not bag-holds. Our recommended template: migration gate at 85% complete, take-profit at 40-80%, stop-loss at 20-30%, and trailing stop at 15%. The idea is to capture the graduation re-rate (often 50-200% within minutes), then let the trailing stop ride it if the move continues. We're publishing the full template in the configure UI — toggle 'Migration Snipe' and the defaults line up.
Is this the same as Axiom or Photon sniper bots?+
Those are generic sniper bots — they fire on any new pump.fun mint matching criteria (volume, dev buy %, holder count). Migration sniping is a DIFFERENT strategy, aimed at late-stage tokens that have survived the first 60 minutes and are now positioned to migrate. The two can live side-by-side: run a generic early sniper on one config, and a migration-gated copy-trade on another. They catch different parts of the curve.
What about tokens that never graduate?+
Most Pump.fun tokens stall before 85%. That's the point of the gate — skipping a buy because curve is at 73% protects you from chasing a stalled coin. The gate is a true filter, not a sort: a token that hits 85% and then reverses back to 60% no longer passes the check, so a later whale re-buy on the same token wouldn't fire. Combine the gate with your anti-rug level and cooldown for extra protection.
When does phase 2 ship (real-time / PumpSwap / graduation-moment)?+
Phase 1 (live today) is a pre-migration gate on top-trader buys with direct-builder execution. Phase 2 adds a Yellowstone gRPC program-subscription on the Pump.fun migration instruction itself (the graduation moment), multi-tx Jito bundles for atomic approve+swap, and a native PumpSwap pool builder for the first-second-after-graduation leg. Phase 3 adds a pre-signed buffer so our transaction is encoded and ready to broadcast the slot we see the migration trigger. See the migration roadmap section below.
Can I combine migration snipe with smart-money confluence?+
Yes — both are independent filter gates. Smart-money confluence requires N tracked wallets to buy the same token within a window, migration snipe requires the token's curve to be ≥ X% complete. Enable both and you only fire when multiple whales rotate into a token that's mathematically close to graduation. That's the highest-conviction tier we support.
Ready to stop chasing graduated candles?
Enable migration snipe on one of your existing copy-trade rules and let the gate do the filtering. Every non-graduating token gets skipped automatically; you only fire on the mints that actually reach the graduation threshold.
Works with fixed-amount, mirror-%, and scaled sizing · Free on the TRADER plan