Your strategy, executed.
Never your private key.
Your Pine strategy fires the alert. The order, the stop and the exit land on Hyperliquid exactly as written — through a permission you can read and revoke.
- Your private key never leaves your wallet.
- Trade-only agent wallet, revocable in one click.
- Exits and stops that fill mid-bar still execute — nothing refused as “stale”.
- Your system, executed — while you live your life.
Every entry, re-entry and exit — pushed to Telegram the moment it fills.
What you actually do
Paste the alert template into your strategy.
Any Pine Script v6 strategy works — the template drops into your alert message and fills itself from your entries and exits.
Connect a trade-only agent wallet.
One signature, scoped to trading only, revocable by you at any time. Your private key never leaves your wallet.
That’s it. Signals execute 24/7.
Every entry, exit, and stop lands on Hyperliquid whether you’re at the screen or asleep — with the receipt on your phone before you’ve looked up.
paste · sign · fill
approveAgent permission before signing · watch a real signal fill on Hyperliquid and the receipt land in Telegram.Under the hood
Pine Script v6
Your strategy decides. Deterministic, backtestable, no AI in the live path.
Authenticated webhook
Every alert carries a signal id, a timestamp, and a shared secret compared in constant time. Duplicates and stale signals die at the gate.
Execution layer
Validated, deduped, freshness-checked, ownership-guarded — then executed. Exits and stop-moves are exempt from the freshness gate on purpose: TradingView composes them at bar close, the fill happens intrabar, and a stop that fires mid-bar must never be refused as “stale”.
On-exchange safety
SL and TP placed as real orders at entry. They survive the server dying.
Don't trust the claim. Read the signature.
Everyone writes "non-custodial" on their landing page. Here's the actual permission you'd be granting — request it on testnet right now, in your browser, no install, and read it yourself before you believe anybody, including me.
Hyperliquid approveAgent
The boring fields that keep you safe
"token": "••••••••", // shared secret — constant-time compare
"symbol": "BTCUSDC.P", // TV ticker, mapped server-side
"action": "buy",
"side_intent": "entry",
"qty_pct": 25,
"leverage": 3,
"sl_price": 64100,
"tp_price": 69800,
"signal_id": "x7f3…", // dedupe — duplicates die at the gate
"ts": 1787654321000, // staleness gate — old signals refused
"strategy": "your-strategy-id" // one strategy owns each coin
}
This is the literal payload the executor accepts — not a simplified mock. Three of these fields do nothing on a good day and everything on a bad one: signal_id kills duplicates, ts refuses stale signals, and on exits reduce_only: true means a close can never accidentally open a position. On exits the size is also clamped to the live position on the exchange — a late or duplicate close can shrink a position, never flip it.
Architecture, not promises.
Built by a developer who writes Pine Script v6 daily — and publishes the failures too: every bug caught, every rejection logged, every week.
Provably non-custodial
Agent wallets grant trade-only rights at the exchange level. Not a policy, not a promise — a permission you can read and revoke yourself.
Exits that survive the bar
TradingView writes an exit alert at bar close; the trailing stop fills mid-bar. A naive executor sees an “old” timestamp and refuses its own exit. Here, exits, reduce-only closes and stop-moves skip the freshness gate, are guarded by the dedupe id instead, and can only shrink a position — never flip it. If a stop can’t be replaced, the position is never left unguarded. Learned in testnet, kept on purpose.
Silence is never "fine"
Every rejection, clamp and refusal pushes an alert and lands in an audit ledger. A system that fails quietly is a system you can't trust.
Deterministic, not an LLM
No AI makes live trading calls. Your Pine strategy decides; the executor carries it out. Same signal in, same order out — every time, backtestable.
I only get paid when your strategy actually trades.
Revenue comes from Hyperliquid builder codes — a native 0.01% fee on executed volume, approved once by you, capped by the exchange, and revocable the same way the agent wallet is. No funds ever route through me: it’s fee attribution by the exchange, not a payment I collect. If the bot never trades, I never earn. Execution stays pay-per-fill. If I ever offer something beyond execution — strategy rental, say — it’ll be priced on its own and announced in the ledger before it exists anywhere else.
Asked and answered.
What happens if your server dies mid-trade?
Nothing happens to your protection — that’s the point of the architecture. Stops and targets are placed as real orders on Hyperliquid the moment a position opens. The exchange holds them, not my server. A dead server means no new trades, never an unguarded position.
What can the agent wallet actually do?
Place and manage orders on the account it’s scoped to. That’s the whole list. It cannot withdraw, transfer, bridge, or touch collateral — not as policy, but because the permission itself doesn’t include it. The testnet demo above lets you read the exact permission before you believe me.
How do I revoke it?
In the Hyperliquid app: API → your agent → revoke. One click, on the exchange, with no involvement from me. Works even if this site disappears tomorrow.
Which TradingView plan do I need?
Any paid plan — webhook alerts aren’t available on the free tier. Your existing strategies don’t need rewriting; the alert template drops into what you already have.
Which wallets work?
Any EVM wallet — MetaMask, Rabby, Rainbow, or a hardware wallet through them. Heads up: wallets may flag this as a new/unlisted site. That’s expected for a domain this young, not a red flag — and if you’d rather never connect your main wallet to any new site (smart habit), a fresh wallet works the same.
Who’s building this?
One developer, in public. I write Pine Script v6 daily, run this exact infrastructure on my own capital — testnet today, mainnet only when the evidence says so — and publish the weekly ledgers, including the ugly weeks. I’m pseudonymous, but everything is checkable: this site’s code is public on GitHub, the execution receipts land on a public Telegram channel as they happen, the weekly ledger lands in your inbox, and the testnet demo above doesn’t care who I am — read the signature instead of trusting a face.
When does mainnet happen?
When the evidence says so, not when marketing does: the gate is clean, fully-organic trade cycles on the current code, logged in public. Testnet receipts are published weekly either way. Waitlist hears first.
What do you store?
No keys, ever — they never leave your wallet. The waitlist stores your email and nothing else.
What happens when my trailing stop moves or fills mid-bar?
It executes. TradingView builds exit alerts at bar close, but a trailing stop fills intrabar — so a naive executor sees an “old” timestamp and refuses its own exit. Here, exits, reduce-only closes and stop-moves are exempt from the freshness gate and guarded by the dedupe id instead. If a stop can’t be replaced, the position is never left unguarded. This is the bug that bit me in testnet and the reason the exemption exists.
Can’t I just trade Hyperliquid from TradingView?
No. TradingView added native Hyperliquid data in July 2026 — you can chart HYPERLIQUID:BTCUSD around the clock — but it’s charting only. There’s no broker integration, so no order ever leaves the chart. A webhook is the bridge. That’s what this is.
Watch it get built, receipts and all.
Weekly ledgers: signals fired, orders executed, rejections caught, bugs found. Good weeks and ugly ones.