Quick start

Walk through one full Bi task loop in five minutes.

This page walks you through one complete loop: ask → the agent works → review the artifacts.

1. Pick a workspace

Choose the project root in the UI. Bi confines the agent’s file operations to that directory; anything outside is rejected.

2. Ask for something specific

Rather than “take a look at this project”, give a goal you can verify:

Read the entry files under src, map the module dependencies, and write them to DEPENDENCIES.md.

A concrete goal makes it easier for the agent to know when it is done.

3. Watch it work

The conversation pane shows tool calls live: which files it read, which commands it ran, what it changed. The first time it writes a file or runs a command, Bi asks for permission.

You can auto-approve low-risk operations like read-only inspection and test commands in settings, so it interrupts you less.

4. Review the artifacts

Generated documents, images and code appear in the artifacts panel, ready to preview. Terminal output and the change list live there too.

5. Ship it

Changes are visible, and key moments get checkpoints. If the agent went wrong, roll back a step and let it retry — rather than starting over.

Core tools — what the agent can actually do
Permissions — tune the boundary to your taste
Checkpoints — let it act with confidence