Skip to content

Quick Start

This guide walks you through AVC’s core loop: initialize a project, take a baseline snapshot, make changes, see the diff, and restore. By the end you’ll know enough to use AVC daily.

Pick any project directory:

Terminal window
cd ~/my-project
avc init

This creates a .avc/ directory with a SQLite database and a default .avcignore file. Re-running is safe; it’s a no-op on already-initialized projects.

If you want the agent integration files set up at the same time:

Terminal window
avc init --skills claude-code,cursor,windsurf
Terminal window
avc snapshot "Baseline"

Output:

Snapshot created: snap-a1b2c3d4
Label: Baseline
Branch: main
Files: 42
Size: 1.2 MB

This snapshot is your “known good” state. You can return to it at any time.

3. Make changes (or let an agent make them)

Section titled “3. Make changes (or let an agent make them)”

Edit some files. For this walkthrough, just modify any file in the project — change a line in your README, add a new file, whatever.

If you’re using an AI agent, point it at the project and ask it to do something. Examples:

  • “Refactor the auth module to use bcrypt”
  • “Add a unit test for the calculator function”
  • “Fix the bug in the payment handler”
Terminal window
avc snapshot "After changes" --agent claude --notes "Refactored auth"

The --agent and --notes flags are optional but useful — they help you remember what each snapshot represents when scrolling through a long list.

Terminal window
avc diff snap-a1b2c3d4 snap-e5f6g7h8

You’ll see a per-file summary:

M src/auth/login.go (+12 -8)
M src/auth/session.go (+5 -2)
A src/auth/bcrypt.go (+34 -0)

For richer output:

Terminal window
avc diff snap-a1b2c3d4 snap-e5f6g7h8 --json

You’ll get the line-level unified diff for every changed file as JSON — feed it into another tool, or render it in the Web UI.

If the changes broke something, restore the baseline:

Terminal window
avc restore snap-a1b2c3d4

You now know the core loop. From here:

  • List all snapshotsavc list
  • See snapshot details + file treeavc info snap-a1b2c3d4
  • Open a graphical browseravc ui then visit http://localhost:3004/
  • Use branches for parallel agent runs → see Concepts → Branches
  • Browse from VSCode → install the VSCode Extension
  • Wire it up to your AI agent → see Agent Integration