Gordon is for traders who understand markets but are bottlenecked by tooling. People with genuine ideas about where prices are headed who don't want to spend weeks writing Python just to test a hypothesis. People who want the power to turn intuition into a risk-sized plan.
You talk to it in natural language. It reasons about markets on your behalf: pulling data, running analysis, constructing trade plans, managing execution. Everything runs locally. Your keys never leave your machine. It doesn't replace your judgment. It extends your discipline.
Vibe Trading
Vibe coding changed how software gets built. Describe what you want, let the machine handle implementation. The programmer becomes a director. Tools like Cursor and Claude Code already work this way.
The same shift is coming to markets.
Express your intent, refine it interactively, execute with confidence. Instead of code files, you're working with positions. Instead of test suites, you're running backtests. Instead of deploying to servers, you're managing live risk.
Say "show me funding rates on SOL perps" and a chart materializes. Say "set up a grid strategy between 140 and 160" and the agent drafts a complete plan: capital requirements, expected fees, risk scenarios if the range breaks. Context becomes something you reference directly: mention @ETH and the system pulls price, volatility, flows, and open interest automatically.
How It Works
Natural language in, market action out. You describe what you want. Gordon translates intent into concrete plans, showing you exactly what will happen before anything executes.
Local-first architecture. Your API keys stay on your hardware. Strategies execute on infrastructure you control. Full sovereignty over your data and your risk.
Human approval on every trade. Before execution, you see a portfolio diff: current exposure, exposure after, slippage, fees, risk impact. You confirm or reject. The system never moves without you.
Discipline becomes structural. Define your constraints upfront: position limits, asset restrictions, stop-loss defaults. Gordon enforces them even when volatility tests your resolve.
The Roadmap
Phase One: Command Line Now
Proving the thesis: an AI agent that reasons about live markets and executes real trades through natural language, safely and reliably. Claude Code for trading.
Phase Two: Generative Interface
Ask for analysis and the right interface spawns dynamically: a chart, a risk visualization, an interactive widget. The interface becomes a byproduct of reasoning.
Phase Three: Full Environment
An IDE for directing financial agents. Multiple specialist agents debating before you commit capital. Cursor for markets.
Under the Hood
Infrastructure built for AI agents as first-class citizens, not awkward additions to legacy systems.
A state layer maintains consistent market reality across the system. A tool registry gives standardized access to any exchange or data source. An event bus enables real-time reaction without polling. A safety layer intercepts every action before it touches the outside world, checking constraints and requiring confirmation. The entire architecture is local-first by design.
Get Early Access
Now in private alpha.
Join the Waitlist