Build a trading agent
that never blinks.
Nightwatch turns a strategy into an autonomous agent — one that reads every tick, reasons about entries and exits, and executes on your behalf, wide awake at 3am when the setup finally shows up.
Three steps. One agent watching the tape.
Describe the edge
Compose conditions and actions in the Agent Console, or describe your setup in plain language. Nightwatch compiles it into an executable strategy graph.
Stress it against history
Every agent runs a walk-forward backtest across tick-level data before it sees a live order — win rate, drawdown, and slippage included.
Send it live, with guardrails
Promote the agent to a connected broker with position limits, a daily loss cap, and a kill switch already wired in.
Everything an agent needs to trade responsibly.
Multi-agent orchestration
Run a book of agents across asset classes, each with its own mandate, sized against a shared capital limit.
Guardrails by default
Max position size, daily loss caps, and a kill switch that flattens positions the instant a limit trips.
One execution layer
Connect equities, futures, and crypto venues once — an agent's logic runs identically across all of them.
Tick-level backtesting
Walk-forward tests replay years of tick data with realistic slippage, so live results shouldn't surprise you.
A reasoning journal
Every decision an agent makes is logged in plain language — what it saw, what it weighed, why it acted.
Always-on monitoring
A live dashboard of every agent's exposure, P&L, and health — with alerts routed to you the moment something looks off.
Build an agent right here.
Pick a condition and an action below. Nightwatch compiles it into a strategy, then run a backtest to see how it would have traded.
// Select a condition and an action // to compile your agent's logic.
Built for agents that hold real capital.
Start free. Pay when an agent goes live.
Starter
- 1 live paper-trading agent
- Tick-level backtesting, 1 year of data
- Community strategy templates
Desk
- Up to 10 concurrent live agents
- Full tick history, 8 years back
- Guardrails, kill switch, alerting
- Priority execution routing
Fund
- Unlimited agents, shared capital limits
- Team roles and audit trail
- Dedicated execution latency SLA
- Direct line to the desk support team
Before you deploy an agent.
No. You compose conditions and actions visually in the Agent Console, and Nightwatch compiles them into an executable strategy. If you want finer control, you can drop into raw Python at any point.
Nightwatch connects to major equities brokers, futures venues, and crypto exchanges through a single execution layer, so one agent definition can trade across every connected venue.
Every agent runs inside guardrails you set before it goes live: a max position size, a daily loss limit, and a kill switch that halts the agent and flattens positions the moment a limit is hit.
Yes. Every agent defaults to a simulated account fed by live market data. You promote it to real capital manually, whenever you decide it's ready.
Agents treat a stale or disconnected feed as a reason to pause, not to guess. A watchdog process halts new orders the moment data quality drops below the threshold you set.
Your next trade doesn't need you awake for it.
Build an agent in minutes, backtest it against real history, and decide for yourself when it earns real capital.