First run

The window

Arvo opens like an editor: an activity bar on the left (Portfolio, Research, Files, Search, Accounts, Extensions, Alerts, Settings), tabs in the middle, a bottom panel (Problems, Output, Terminal, Jobs) and, when you open it, the Agent panel on the right. Everything is a command: the menus, the keybindings and the palette (Ctrl+Shift+P) are three views of one registry, with VS Code's names where VS Code has one.

The Welcome tab lists the tabs worth knowing: Rulesets, Risk, Sessions, Portfolio, Runs, Operations, Agent.

The project folder

Arvo works in a folder. Choose one on the first run (File → Open Folder); Arvo remembers it. Everything the engine produces lands there — the bar library, findings, session records — as files you can read, copy and put under git. Your own scripts live there too.

The engine

The window starts an engine for you and shows its state in the status bar. The engine writes engine.json (its address and the research token) and control.json (the control token) to the app data directory, which is how every other client — a script, an agent, the command line — finds it.

Closing the window does not stop the engine: it minimises to the tray and keeps serving, so a schedule runs and a session trades with the window closed. Quit Arvo from the tray stops both. Halt trading from the tray is the kill switch.

Your first study

1. Open Research in the activity bar, pick a source (Alpaca, Robinhood or Yahoo, whichever you connected) and fetch bars for an instrument.
2. Pick a rule (start with Moving-average crossover — it is the control, a rule nobody disputes) and run a study.
3. Read the verdict and the advice above the numbers. Most first studies are Inconclusive; that is the platform working. How to read one →

Data you can try without an account

The library ships with three simulated instruments — DRIFT.SIM, TREND.SIM and NOISE.SIM — so a study can be run before any vendor is connected. A rule that "works" on NOISE.SIM is the first thing to be suspicious of.