Skip to main content

Use With ChatGPT, Claude, Codex, Cursor, and Other LLM Tools

PastePrompt does not include LLM access, LLM API keys, or a hosted model. It prepares local context bundles that you decide to copy, export, paste, attach, or import into another tool.

General paste/export workflow​

  1. Build a focused bundle in PastePrompt.
  2. Check token estimates.
  3. Run the secret scanner.
  4. Review the generated output.
  5. Copy XML-like context or export Markdown.
  6. Paste, attach, or import the bundle into your chosen tool.
  7. Ask a narrow question.
  8. Validate the response against local source code and tests.

Model context limits​

Different tools and models have different context windows and token accounting. PastePrompt token counts are estimates, so leave margin.

If a bundle is too large:

  1. Remove generated or low-value files.
  2. Split by module or review question.
  3. Export one bundle for source files and another for tests or docs.
  4. Use Git diff context for patch review instead of full repository context.
  5. Summarize previous model output manually before sending the next bundle.

ChatGPT​

Use copy or Markdown export depending on the workflow:

  • Paste XML-like context for a structured chat prompt.
  • Attach or paste Markdown when you want a readable artifact.
  • Ask focused questions and request code-location evidence.

Do not assume ChatGPT has access to files outside the bundle unless you provide them separately.

Claude​

Use the same bundle discipline:

  • Keep instructions precise.
  • Split large repositories into smaller modules.
  • Ask the model to distinguish code facts, assumptions, and hypotheses.

Do not rely on provider-specific context handling without checking the current tool behavior.

Codex​

For Codex-style coding agents:

  • Use smaller bundles with exact task instructions.
  • Include files that define the behavior under review.
  • Include tests when asking for validation or patch suggestions.
  • Keep source-of-truth files local and verify suggested changes yourself.

PastePrompt bundles can help orient an agent, but they do not replace repository access or local tests.

Cursor​

Cursor workflows often pair well with editor handoff:

  1. Open relevant files from PastePrompt in Cursor when needed.
  2. Copy or export a focused context bundle.
  3. Use Cursor's local editor context alongside the PastePrompt bundle.
  4. Rebuild the bundle if the selected files change.

Local LLMs​

For local models:

  • Use Markdown or XML-like exports as plain text input.
  • Keep bundles smaller if the model has a limited context window.
  • Store exports in a local folder you control.
  • Confirm the local tool does not sync prompts or logs unexpectedly.

Privacy boundary​

PastePrompt prepares the bundle locally. External tools are separate. After you paste, attach, or import a bundle, that tool's privacy settings, terms, logs, and retention behavior apply.