Skip to main content
When you give Julius a task, it takes actions, not only text replies. Each action is a tool: running code, querying a database, searching the web, building a slide deck. You’ll see each step labeled in the chat as it happens, like Generated code, Searched internet, or Created slides. Expand any step to see what Julius did. You never invoke these yourself. Julius decides which tools to use based on your request, your model, your plan, and the data you’ve connected.
Tools also drive your credit usage. Running code, searching the web, or generating a deck costs more than a simple text answer.

​
Core tools

​
Run code

The tool behind most of what Julius does. Julius writes Python or R and runs it in a secure cloud container to clean data, run analysis, build models, and produce charts. Code runs automatically, and Julius installs any libraries it needs. In the chat this appears as Generating code → Generated code. You can read the code, copy it, and see its output inline.
Clean this dataset and show me the correlation between spend and revenue

​
Edit files (apply patch)

When Julius works with files or code, it can make surgical edits instead of rewriting everything, changing only the lines that need to change. Long files and scripts stay intact as it iterates. In the chat this appears as Editing file → Edited file, with a diff of what changed.

​
Build visualizations

Julius generates interactive charts from your data (bar, line, scatter, and more) that you can hover and read in the browser. For static charts, see Data visualizations. For full dashboards and apps, see Interactive artifacts. In the chat this appears as Generating visualization → Generated visualization.

​
Create slides

Julius builds a full slide deck and renders it in the browser, with one-click export to PowerPoint. Use it when you want a deliverable to present. In the chat this appears as Creating slides → Created slides. See Slides for prompting tips and export details.
Turn this analysis into a 10-slide deck for my team

​
Generate images

Julius can create and edit images from a description: diagrams, illustrations, or visuals for a deck or report. In the chat this appears as Generating image → Generated image.

​
Working with your data

​
Query connected databases

When you’ve connected a database or warehouse, Julius writes and runs SQL against it directly. That covers Postgres, MySQL, SQL Server, Snowflake, BigQuery, Databricks, and more. You’ll see steps like Querying BigQuery → Queried BigQuery. See Data Connectors.

​
Explore your schema

Before querying an unfamiliar database, Julius can search the schema to find the right tables and columns, so it writes correct SQL instead of guessing. This shows up as Searching schema and Looked up schema.

​
Pull financial and market data

For finance questions, Julius can pull market and financial datasets (prices, fundamentals, filings, and macro data) and analyze them like any other source. This appears as Querying financial data → Queried financial data.

​
Research & assistance tools

​
Search the web

Julius can search the internet to pull in current information your data doesn’t contain, like context, definitions, or recent events. This appears as Searching internet → Searched internet.

​
Write reports

Julius can produce a long-form, structured report from your conversation and revise it section by section. See Reports.

​
Ask you a question

When a request is ambiguous, Julius may pause to ask you a quick clarifying question rather than guess, so the work matches what you wanted.

​
Connect external tools (MCP)

If you’ve connected tools through the Model Context Protocol, Julius can call them as part of a task. See MCP.
The available tools vary by model, plan, and the data sources you’ve connected. The list above is what Julius can do, not a fixed menu shown every time.
Wondering whether Julius can do something specific? Ask it in the chat, or email team@julius.ai.