Vector is an AI coding agent for VS Code. It helps you understand an existing codebase, plan a change, edit files, run commands, and validate results while you remain in control of every consequential action.
Work with your codebase
Start with a goal, a bug report, or a screenshot. Vector can inspect relevant files, search the workspace, follow project instructions, and build a focused working context before proposing or making changes.
- Create and edit files with reviewable diffs.
- Run terminal commands after your approval and use their output to continue debugging.
- Use browser automation to exercise local web applications and capture runtime evidence.
- Add context from files, folders, terminals, diagnostics, Git changes, and URLs.
- Extend workflows with Model Context Protocol (MCP) servers.
Choose how Vector works
Use Plan mode to explore options and agree on an approach before changing files. Use Act mode when you are ready for implementation. Vector asks for permission before executing commands or modifying your workspace, so you can inspect each step.
Models and accounts
Sign in to Vector to access the current Vector model catalog. Recommended and free model cards are populated from the live service; unavailable preview entries cannot be selected or used for inference. You can also configure supported external or local model providers from Settings.
Get started
- Install the extension and open a folder in VS Code.
- Open the Vector view from the Activity Bar.
- Sign in, select a live model, and describe the outcome you want.
- Review Vector's plan, proposed edits, and command requests as it works.
License
Apache-2.0