Core Concepts

Canvas & Elements

The canvas is an infinite 2D workspace. You place elements on it: frames (interactive visuals), notes (markdown text), PDFs, images, and embeds (external URLs). Every element can be dragged, resized, and connected with edges.

Frames

Frames are the workhorse element. Each frame is a self-contained mini web page. The agent writes custom code for every frame, including Plotly charts, D3 visualizations, data tables, and dashboards. Frames can load local assets (images, data files) and interact with the data engine.

Skills

Skills are specialized analysis capabilities. The agent has 56 skills spanning scoring, valuation, sector analysis, portfolio management, screening, and more. When you ask for a DCF model or a sector deep-dive, the agent activates the relevant skill and loads a detailed execution plan to guide the analysis.

Projects

Projects group related work. A project has its own canvases, data, notes, and deployment config. Use projects to organize research by company, sector, or theme.

Refresh Bindings

A refresh binding connects a Python script to a canvas element. On a timer (e.g., every 5 minutes), the script runs, fetches fresh data, and pushes the result to the frame. Your dashboards stay live without manual intervention.

Recipes

Recipes are saved analysis workflows. If you frequently run the same type of analysis (e.g., “quarterly earnings review for my watchlist”), save it as a recipe. Re-run it anytime with one command.

Data Engine

Jetro runs a local DuckDB database. API responses are cached automatically, imported datasets are registered as tables, and you can query everything with SQL. The agent can also write SQL for you.