How CodeZero works
CodeZero is an AI engineering workspace inside VS Code that helps developers and coding agents understand existing software before changing it. It brings repository structure, system relationships, product features, implementation context, engineering plans and Git history into the same workspace — including context from repositories that are indexed by your organisation but are not checked out locally.
What CodeZero provides
AI can generate code quickly. The harder part is understanding the system around the requested change.
Before implementation begins, developers still need to answer questions such as:
- Where is this behaviour implemented?
- Which repositories are involved?
- What depends on the component being changed?
- Does a similar implementation already exist?
- What else could be affected by the change?
CodeZero builds and maintains this context so it does not need to be reconstructed for every new task.
Where CodeZero fits
CodeZero runs inside VS Code and works alongside the coding agent you already use. It does not replace Claude Code or Codex or Microsoft Co-Pilot and does not modify its credentials — it continues to use whichever subscription or API credentials you have configured for it.
CodeZero owns the context layer: the organisation's repository index, the relationships between components, the feature maps and the plan under review. The agent owns execution.
Repository and Git-oriented views work without an agent. Agent turns and AI-assisted feature naming require a compatible Claude Code installation.
Understand → Plan → Build
CodeZero follows a simple engineering workflow:
- Understand
- Explore the system, trace a feature, ask questions and inspect the history behind the code.
- Plan
- Describe the change, clarify decisions and review the proposed implementation before execution.
- Build
- Let the coding agent execute with the relevant repositories, dependencies and approved scope already available.
Explore the workspace
The CodeZero workspace includes:
- Code Graph for understanding how the system is connected
- Functions for Feature Mapping and tracing product behaviour
- Context-aware Chat for investigating the system and requesting changes
- Plans and Tasks for reviewing and structuring implementation
- Git Tree for repository history and everyday Git operations
- Auto, Planned and Normal modes for controlling how changes are handled
- Coffee Mode for finishing approved work while you step away
The following guides cover each part of this workflow.