Context Injection
Four methods to inject context — auto-stack, AI config files, clipboard, and file attachments.
Context Injection lets you provide additional context to Vocoding beyond what is in your project configuration. This context is included in your prompts to produce more relevant results.
4 Methods to Inject Context
1. Auto-Stack Detection (Automatic)
When you set a project root in Project Context (Cmd + 3), Vocoding automatically detects your tech stack and includes it in prompts.
No action needed -- just set up your project profile once. Vocoding reads marker files like package.json, Cargo.toml, and requirements.txt to determine your stack.
2. AI Config Files (Automatic)
Files like CLAUDE.md, .cursorrules, and conventions.md in your project root are automatically included when enabled.
| File | Description |
|---|---|
CLAUDE.md / .claude.md | Claude Code instructions |
.cursorrules | Cursor IDE rules |
.codex.md | Codex instructions |
.gemini.md | Gemini instructions |
conventions.md | Project conventions |
.copilot-instructions | GitHub Copilot rules |
Toggle individually in the right panel > Context tab, or in Project Context settings.
3. Clipboard Injection (Manual, Instant)
Copy code or text to your clipboard, then press the context injection shortcut:
| Platform | Shortcut |
|---|---|
| macOS | Cmd + Shift + Space |
| Windows | Ctrl + Shift + K |
The clipboard content appears as a context block in the composer -- a dark card showing a preview of the content, with edit and remove buttons.
Enable/Disable: Settings > Context Injection toggle
Use Cases for Clipboard Injection
- Copy a code snippet > inject > ask about it
- Copy an error message > inject > get debugging help
- Copy API documentation > inject > generate implementation
- Copy a design spec > inject > generate matching code
4. File Attachments (Manual)
Drag and drop files onto the composer, or click the paperclip icon to browse.
| Limit | Value |
|---|---|
| Max files per message | 5 |
| Max size per file | 1 MB |
| Supported types | txt, md, json, and other text formats |
Files appear as attachment chips above the input area.
Best Practices
- Use clipboard injection for quick context -- it is the fastest way to add a code snippet or error message
- Use file attachments for larger documents -- style guides, API specs, example outputs
- Keep AI config files concise -- they are included in every prompt, so verbosity costs tokens
- Disable unused context sources -- don't waste token budget on irrelevant files
- Check the Context tab in the right panel to see exactly what is being sent to the LLM