Building a Prompt Library with Lexicon

If you use Claude Code, Claude Desktop, or any AI assistant daily, you have probably written the same prompt dozens of times. Your code review instructions. Your writing style guide. That specific way you like technical documentation formatted. Each time, you retype it from memory or scroll through old conversations trying to find the version that worked best.

Some people save prompts in text files or Notion pages, but that creates its own friction. You have to find the file, copy the text, paste it into your AI tool, and manually substitute any variables. There is no search, no organization, no way to track which prompts you actually use versus the ones you saved once and forgot about.

The real cost is not just time -- it is consistency. When you retype prompts from memory, they drift. The careful instructions you refined over weeks gradually lose their precision. And when you discover a better approach, there is no way to update it everywhere at once.

Lexicon is a native macOS prompt library with built-in MCP integration. Your prompts become tools that Claude Code and Claude Desktop can access directly. Instead of copying and pasting, you invoke prompts as slash commands. Instead of scattered text files, you have a searchable, categorized library with template variables and smart collections.

Like all Helton Labs products, Lexicon is local-first. Your prompt library lives in SQLite on your Mac. No cloud accounts, no syncing to someone else's servers, no worrying about whether your proprietary prompts are being used as training data.

Organize prompts with categories, tags, and template variables

Organize by Category and Tags

Structure your prompt library the way your brain works. Create categories for code review, writing, analysis, debugging, architecture, and whatever else matches your workflow. Add tags for language-specific prompts, project-specific contexts, or difficulty levels. Lexicon's information-dense table view handles libraries of any size with instant search and filtering.

Template Variables

The real power of a prompt library is reusability, and that requires variables. Lexicon supports {{variable}} placeholders with defined types, defaults, and validation. A single "Code Review" prompt can work for any language by using {{language}} and {{codebase}} variables. Preview the rendered output before sending to make sure substitution looks right.

MCP Integration

Lexicon ships with a built-in MCP server. Once connected to Claude Code or Claude Desktop, your entire prompt library is available as tool calls. Type /lexicon: and get autocomplete across all your saved prompts. No copying, no pasting, no context switching. The prompt goes directly from your library into the conversation with all variables resolved.

Compose with Other Tools

Because MCP connects tools together, Lexicon prompts can reference other MCP-enabled apps. Create a "Tech Debt Audit" prompt that instructs Claude to scan your codebase and then create ContextSwitch tasks for each issue found. Or a "Client Check-in" prompt that pulls recent interactions from Evergreen and drafts follow-up emails. Your prompts become workflows that span multiple tools.

Smart Collections

Lexicon tracks how often you use each prompt and when you last used it. Smart collections auto-update based on these signals -- see your most-used prompts, recently added prompts, or prompts you have not touched in months that might need updating. Manual collections let you curate specific sets for different projects or contexts.

Import and Export

Share your best prompts with your team by exporting to JSON. Import prompt libraries from colleagues or community collections. Batch operations make it easy to move entire categories between libraries or merge collections after trying a new set of prompts.

Lexicon is for anyone who uses AI tools daily and wants to get more consistent, efficient results. Developers using Claude Code for code review, debugging, and architecture. Writers who need consistent style and formatting instructions. Analysts who run the same types of queries regularly. Anyone who has realized that the quality of AI output depends heavily on the quality of your input -- and that good prompts are worth preserving, organizing, and refining over time.

Learn More About Lexicon