GeezerVison

CodebaseMD is a VS Code extension that exports a workspace, a selection or your Git changes as one Markdown, XML or plain-text document, ready to paste into an AI assistant or a review.

Download · Documentation · Changelog · Contributing · Security · Support

Status: 4.0.0 is the current release and requires VS Code 1.146 or later. On an older VS Code, or a fork based on one, use 3.0.0, which runs on 1.70 and later. See the changelog and the limitations.

• What CodebaseMD is
• A short example
• Install
• Commands
• Settings
• Choosing what is shared
• How it works
• Limitations
• Repository layout
• Development
• Documentation
• Contributing
• License

Giving an assistant or a reviewer the right code usually means copying files one at a time. CodebaseMD does it in one step: pick a scope, and it writes a single document with a folder tree, per-language statistics, a token estimate and the contents of each file.

• Scopes: the workspace, selected files and folders, changes against HEAD, or the current file (including unsaved edits).
• Formats: Markdown with a linked contents list, escaped XML, or plain text.
• Destinations: a file, the clipboard, or a new editor tab.
• Content modes: full source, a compact Micro summary, or a declaration skeleton, for fitting more files into a context budget.
• Guard rails: recognizable secrets are redacted by default, ignore rules are respected, symbolic links are not followed, and a warning appears before an export exceeds your token budget.

CodebaseMD runs entirely on your machine. It does not upload source code, call an AI provider or collect telemetry.

├── .env
├── README.md
└── src/
    └── greet.ts

API_TOKEN="sk_live_abc…"

# Project Export

3 files | 28 estimated content tokens (characters / 4) | 1 redactions | full mode

## Languages

| Language | Files | Estimated tokens |
| --- | ---: | ---: |
| dotenv | 1 | 7 |
| markdown | 1 | 2 |
| typescript | 1 | 19 |

## Contents

- [\.env](#file-1) — dotenv, ~7 tokens
- [README\.md](#file-2) — markdown, ~2 tokens
- [src/greet\.ts](#file-3) — typescript, ~19 tokens

### \.env

```dotenv
API_TOKEN="***REDACTED***"
```

### src/greet\.ts

```typescript
export function greet(name: string): string {
  return `Hello, ${name}`;
}
```

The output above is abridged: a real export also lists the largest files and the folder tree. XML and plain-text exports carry the same information.

1. Download the .vsix from GitHub Releases. A SHA-256 checksum is published next to it.
2. In VS Code, open Extensions → … → Install from VSIX… and select the file.
3. Open a local folder and run CodebaseMD: Export Wizard from the Command Palette.

The extension is also on the Visual Studio Marketplace. Marketplace publication is a separate step and can lag behind GitHub Releases. Git is needed only for Git exports. See installation for details.

The two selected-export commands also appear in the Explorer context menu. Multi-root workspaces prompt for one folder per export. Usage walks through each command.

Search for CodebaseMD in Settings. Every key has the codebaseMD. prefix and can be set per workspace folder.

The wizard starts from these values. The four original commands always save Markdown in their full or Micro mode, and Copy Current File always uses the clipboard. Configuration has examples.

The same rules apply to explicit selections and to Git exports.

VS Code command ──► config.ts ──► exporter.ts ──► formatters.ts ──► file / clipboard / editor
   (extension.ts)   validated      │
                    settings       ├─ fsUtils.ts   ignore rules, containment, directory walk
                                   ├─ git.ts       changed files and filtered diffs
                                   ├─ secrets.ts   redaction, before any transformation
                                   ├─ analyzer.ts  Micro summaries, language mapping
                                   ├─ skeleton.ts  declaration outlines
                                   └─ tokens.ts    characters / 4 estimate

Only extension.ts and config.ts touch VS Code APIs. The export pipeline is plain Node code, which is why the unit tests can run it against temporary directories and Git repositories without an editor. Nothing is written until the export has been prepared and the budget check has passed. Architecture describes each stage and the safety properties it keeps.

• Token estimates are ceil(characters / 4), not a model tokenizer. Real usage can differ substantially.
• Secret redaction is heuristic. It recognizes common provider tokens, JWTs, private keys and password or token assignments; it can miss secrets and can redact ordinary values. Review exports before sharing them. Paths stay visible unless excluded.
• Micro and skeleton modes use regular expressions, not language parsers. They can miss multiline declarations or misread signatures. Use full mode when exact source matters.
• Git exports reflect the disk and index against HEAD. Unsaved buffers do not select a file, diffs do not include untracked files, and Git exports require a trusted workspace.
• Workspaces must be local folders, one folder per export. Virtual filesystems and untitled files are not supported.
• Platforms: CI covers Linux and Windows. macOS is expected to work but is not tested in CI.

You need Bun 1.4+, Node.js 22.12+ (24 recommended) and Git.

bun install --frozen-lockfile
bun run check              # Biome lint and format check, then unit tests
bun run test:integration   # extension-host tests (Linux: prefix with xvfb-run -a)
bun run package            # build the VSIX

CI runs the same steps on Linux and Windows, audits dependencies, checks REUSE licensing metadata and keeps the built VSIX as an artifact. See CONTRIBUTING.md for the workflow and AGENTS.md for a short guide aimed at coding agents.

Bug reports, fixes and documentation improvements are welcome. Start with CONTRIBUTING.md; use the issue templates for bugs and ideas, and SECURITY.md for vulnerabilities. Everyone taking part is expected to follow the Code of Conduct.

MIT © Alphin Tom. The repository follows the REUSE specification: source files carry SPDX headers and REUSE.toml covers the rest. The Code of Conduct is adapted from the Contributor Covenant under CC BY 4.0.