PT Loopers — v3 (free, EOAbonly, self custodial Pendle PT looping)

Support / Donations

If it saved you money, a tip is appreciated but never required, and it buys you nothing extra.

Chain	Address
BTC	bc1pqve32dpz9zma2l67ns4z0q7w42usdtdgqhwd5pyqng8kv9cacwyszd4xsp
ETH (any ETH token — USDT/USDC/ETH/etc.)	0x523bB94BFe5cf47087eB5556D5ed00515cdC487e
SOL	8UDYmi9Fw7Kg8p8K1xTc2pnViPswVga9vHmTvgwqFBzw
SUI	0xb9726c6024e577684eaeb4670ea97a387b4de558558c3694a62da6179a5ac121

Another method of support/donation is you can trade on https://www.sigmaengine.io to farm multiple perp DEX's with market making/HFT/Delta Neutral strats and more.

Three independent loopers, same strategy (leveraged Pendle PT), different flash/borrow source:

File	Flash/borrow source	Extra cost	Notes
MorphoPTLooper.sol	Morpho Blue flash	0 fee	preferred when Morpho has loan-token liquidity
AavePTLooper.sol	Aave V3 flash	~5 bps premium / leg	fallback for large flashes; needs viaIR
EulerPTLooper.sol	Euler EVC deferred checks	0 flash fee	one Euler market at a time; optional 4626 wrapper

Every user deploys their own looper from the factory and is its immutable owner. No protocol fee exists anywhere in these contracts — there is no treasury, admin, or skim to remove. "Free" means no fee taken from you(the user) by some protocol, but users still pay gas + the borrow interest inherent to leverage (+ Aave's premium on that variant only if you flash via them).

---

updates:

• Slippage (M1): open() now takes an absolute minPtOut (PT units) instead of a bps figure derived from the loan token amount. The old (amount * bps)/10_000 assumed 1 loan-token = 1 PT, which is false on any SY whose exchange rate ≠ 1 (most yield-bearing PTs) — it silently became a no op. Your UI now computes from the Pendle quote.
• EOA-only: createLooper() and open()/close() require msg.sender == tx.origin. This excludes Safe / 4337 / smart contract wallets by design. Remove the onlyEOA modifier + the factory check if you ever want to support them.
• Euler controller release (M3): close() now releases the controller even on a zero debt (unlevered) position, so it can't strand and block the next market.
• Euler YT pre-check (M4): uses convertToAssets(shares) for the exact PT amount, and now also checks YT allowance, not just balance.
• All loopers: close() checks YT allowance up front (clear error instead of a mid flash revert).

Build status: all three compile clean on solc 0.8.19, 0 warnings (Morpho/Euler legacy pipeline; Aave under via_ir = true, set in foundry.toml). Sizes are well under the 24,576-byte limit.

---

Layout & test

./    MorphoPTLooper.sol  AavePTLooper.sol  EulerPTLooper.sol  foundry.toml
test/ MorphoPTLooper.t.sol  AavePTLooper.t.sol  EulerPTLooper.t.sol

(Contracts at repo root, tests in test/ — matching the import "../MorphoPTLooper.sol" paths and src = "." in foundry.toml. Move to src/ if you prefer; just update the imports and src.)

forge install foundry-rs/forge-std
# fill the CONFIG block in each test with a real market, then:
forge test --fork-url $ETH_RPC_URL -vvvv

These have been tested via foundry, compiles + works as intended, however still recommend to run tests with forge test --profile test (test build needs optimizer_runs=1 for via-IR stack; deploy stays at 200) or you get a stack-too-deep error.

---

Deploy

One factory per chain. Fill the canonical addresses (verify each on the explorer):

• Pendle Router V4: 0x888888888889758F76e7103c6CbF23ABbF58F946 (same on every chain)
• Morpho Blue singleton: per chain
• Aave V3 Pool: per chain (Aave variant)
• Euler EVC: per chain (Euler variant)

Then each user calls factory.createLooper() once and operates their looper directly.

---

What the frontend MUST do (this is where user protection lives)

There is no on-chain market allowlist — deliberately. Baking one in would add a trusted admin and break "works with every present/future Pendle market," both of which cut against the free/permissionless model. A clone phishing-UI couldn't abuse the permissionless contract any more than it could abuse a plain token approval, so the contract stays universal and the curation lives in your UI:

1. Compute minPtOut from the Pendle SDK/quote (expected PT out × (1 − tolerance)). Never pass 1 in production — that disables slippage protection.
2. Curate the market list. Only surface (YT, market) pairs you've vetted for a sane oracle and real liquidity. The contract validates the PT/SY/loan-token triplet, but it cannot judge oracle quality or liquidity.
3. Show health / liquidation. These are leveraged positions; if the PT oracle drops or interest accrues past LLTV they get liquidated. Surface the health factor and warn. Use each looper's getPosition(...).
4. Guard the YT. Pre-expiry par-close requires the user still holds the YT (the looper pulls it back). If a user sells/moves it, par-close is impossible and recovery means manually selling PT via the raw execute() escape hatch — expert-only. Warn before any action that would move the YT, and don't construct execute() calldata for non-expert users except the documented unwind.
5. Size leverage against the oracle, not just LLTV. Max borrow ≈ LLTV × (initial + flash) × ptPrice. The naive LLTV/(1−LLTV) bound ignores the PT discount and will revert on discounted PTs.

---

Per-variant gotchas

• Aave: the 5 bps premium is financed as Morpho debt on open and must be covered by redemption on close, leave headroom or a near-liquidation close can revert.
• Euler: one controller (debt vault) per account at a time — fully close before switching markets, or deploy another looper (free). For wrapped-debt markets, fork-test that the wrapper's deposit AND redeem are both open before listing it.
• Morpho/Aave: _debtAssets hardcodes Morpho's virtual constants (1 / 1e6) and the Morpho variant assumes a 0 flash fee — both correct on mainnet today; if either ever changes, open/close revert (fail-safe, not a loss).

---

Still required before real money

1. Run the fork tests above against the exact markets you'll list.
2. External audit. These are immutable and un-pausable; the only recovery path is each user's own execute()