Your own AI. Our infrastructure.
Bring your own strategy rules and your own LLM keys. Run them on the same production rails Tradewink's managed agent trades on — encrypted key custody, layered circuit breakers, most-conservative-wins sizing, slippage-measured execution, and watchdogs designed to treat silence as failure.
Join the Studio waitlist
Early access opens in waitlist order, paper trading first. Joining adds you to the Tradewink mailing list — unsubscribe anytime.
You're right about the LLM part
If you can code, you have seen the pitch: point a frontier model at the market and let it trade. You have probably also read the READMEs. The open-source trading-agent repos you have starred almost all ship the same line — for research purposes only, not financial advice. That is not lawyer boilerplate. It is an engineering confession: getting a model to emit a plausible trading decision is the easy part. Running that decision safely against a real brokerage account, every market day, unattended, is a different discipline entirely.
The LLM writes code, not edge. What separates a running strategy from a burned account is everything around the model call: key custody that does not leak, sizing that refuses to blow up, circuit breakers that end a bad day early, execution that measures its own slippage, and a process that notices when a loop dies silently at 9:31 on a Tuesday. We did not build a smarter model. We built the production layer those READMEs disclaim — and with Studio, we are opening it to your code instead of arguing with it.
The division of labor
You bring
- Your strategy rules — entries, exits, universe, and the logic between them
- Your own LLM keys — your model choice, your prompts, billed to your provider account
- Your brokerage account — funds never touch Tradewink
- Your risk appetite, expressed inside hard platform limits
We run
- Encrypted key custody (Fernet + PBKDF2) and revocation
- Order routing across 13 broker integrations
- A layered risk pipeline of 29 independent checks on every order
- Most-conservative-wins position sizing
- Post-fill slippage measurement (opt-in adverse-selection response)
- Watchdogs, per-tick loop timeouts (rollout in progress), and dead-man health checks
- Walk-forward and Deflated Sharpe validation tooling
- An audit log of every decision from risk check to broker submission
The rails
These rails back Tradewink's managed agent in production. Core protections run by default; the stricter gates are opt-in, off until enabled, and labeled as such below. Studio points the same machinery at your strategy logic.
Encrypted key custody
Broker and LLM API keys are encrypted at rest with Fernet, keys derived via PBKDF2 — the same path Tradewink users' broker keys already go through. Non-custodial: your funds stay at your broker, and keys are revocable at any time.
Layered circuit breakers
A daily-loss breaker halts new entries by default. Two stricter gates — consecutive-loss and daily profit-lock — are available as opt-in additions, off until you enable them. All three only block new entries: protective exits stay allowed, so a tripped breaker never traps you in a position.
Most-conservative-wins sizing
Every position is sized three ways — fixed-risk, ATR-based, and half-Kelly — and the smallest number wins. Your strategy proposes; the sizer disposes.
Slippage-measured execution
After a limit order fills, the executor re-checks the mid 30 to 300 seconds later (tuned per venue) and records the drift — that measurement runs on every limit fill. The automatic response, widening limit pricing when average drift crosses -10 bps (the signature of getting picked off), is an opt-in gate that is off by default.
Watchdogs and dead-man checks
Built on the principle that silence should be treated as failure, not health: watchdogs restart dead processes, hard per-tick timeouts catch hung calls on the agent loops migrated to the new loop framework (a rollout still in progress), and health checks are designed to alert on missing output.
Walk-forward and Deflated Sharpe validation
Overfitting is the default outcome of strategy research, so the tooling assumes it: out-of-sample walk-forward testing, plus Sharpe ratios deflated for the number of configurations you tried before this one.
Also on the rails: 40+ concurrent market-monitoring loops (price, news, SEC filings, options flow), 13 broker integrations, and an audit log on every order. Read the full architecture or the security overview.
What Studio is — and is not
Studio is
- In development, gated behind this waitlist — early access opens in waitlist order.
- Paper mode first. Every Studio strategy starts on paper, the same default as the rest of Tradewink.
- A custom counterpart to our managed pipeline: same rails, your logic in the strategy seat.
Studio is not
- A performance promise. We publish no return projections for Studio because none would be honest.
- Investment advice. Tradewink is a software publisher, not a registered investment adviser or broker-dealer.
- Custody. Your funds and positions stay at your broker; your keys are encrypted and revocable.
Frequently asked questions
What is Tradewink Studio?
Studio is an upcoming way to run your own trading logic on Tradewink's infrastructure. You bring the strategy rules and your own LLM API keys; Tradewink supplies the rails our managed agent already trades on — encrypted key custody, order routing, a layered risk pipeline, conservative position sizing, slippage measurement, watchdogs, and validation tooling. It is in development and gated behind a waitlist.
Do you hold my funds or my keys?
No funds, ever — Tradewink is non-custodial. Your account stays at your broker, and orders route through your own broker API keys. Keys (broker and LLM) are encrypted at rest with Fernet using PBKDF2-derived keys, and you can revoke them at any time.
Can I use my own LLM and API keys?
Yes — that is the point of Studio. Your model choice, your prompts, your API keys, billed to your own provider account. Tradewink wraps the model call in the guardrails that decide whether, how much, and how a resulting trade actually executes.
Will my strategy make money?
Nobody can promise that, and we won't. Most DIY trading systems fail on operations and discipline — unhandled errors, oversized positions, unmonitored processes, overfit backtests — before the market ever gets a chance to judge the idea. Studio is designed to mitigate those operational failure modes; no system catches every failure, and none of it removes market risk. Every Studio strategy starts in paper mode, and nothing here is investment advice.
When does Studio launch?
There is no public date yet. Early access opens in waitlist order, paper trading first. Joining the waitlist costs nothing and adds you to the Tradewink mailing list, which you can leave at any time.
The disclaimer ends here
Your rules, your models, your broker — our custody, sizing, breakers, and watchdogs. Leave an email and we will write to you when Studio opens. Unsubscribe anytime.
Tradewink is not a registered investment adviser, broker-dealer, or financial planner. All data, signals, and analytics on this page are for informational purposes only and do not constitute investment advice, financial advice, or a recommendation to buy or sell any security.
Past performance does not guarantee future results. Trading involves substantial risk of loss, including the possibility of losing more than your initial investment. You are solely responsible for your own trading decisions.