A self-hosted trading desk run by the AI coding agents you already use. Every broker in one account; every trade a commit you approve.


§ I — Why a coding agent
OpenAlice turns trading into a coding task — a workspace, files, version control — so the agent CLI you already pay for can research, decide, and execute. Bring your own; swap it anytime.

§ II — Workspaces
Spin up a workspace for analysis, for trading, for tracking — each its own environment and accountable agent Sessions. Schedule the same analyst or a fresh worker; every report keeps the trail back to who produced it.

§ III — Tracked intelligence
OpenAlice agents can recognize assets, themes, entities, and reports as they work, then create the missing records and link everything back to the thesis that produced it.
Tickers, sectors, topics, reports, and watch items.
Persistent entities the agent can reuse across sessions.
Reports, inbox updates, decisions, and assets in one trail.

§ IV — Capabilities
Six primitives that turn an LLM into something that can actually hold a portfolio without setting it on fire.
Every broker, one ledger.
Stocks, crypto, forex, options, futures, ETFs.
Macro, movers, calendars — no API keys out of the box. Your own keys stay the fallback.
A Python-subset calculator over any source's K-lines — broker or vendor, one call.
Self-scheduling, headless runs, snapshots — Alice never sleeps.
A trading desk used to mean a building, a Bloomberg, and twelve people in shirts.
Now it's a folder on your machine.

Your turn
Open the demo to see the running desk, then pick the install path that matches your machine when you are ready to wire Alice into your agents and accounts.
Need help choosing a path? ask in Discord.