llms.txt Optimization

Mintlify automatically generates llms.txt from your documentation. This guide focuses on writing TSDoc comments that produce helpful descriptions for your generated API reference.For general llms.txt information, see Mintlify’s llms.txt documentation.

Why llms.txt matters for mint-tsdocs

When you generate 400+ API reference pages, llms.txt acts as a table of contents for AI tools. It contains:

Page titles (your class/interface names)
Descriptions (from your TSDoc @remarks)
Organization (from your docs.json structure)

Good descriptions help AI tools:

Find relevant classes: “What caching options exist?” → Finds CacheManager with description
Understand purpose: Knows CacheManager handles caching, not cache instances
Navigate hierarchy: Understands relationships between classes

How mint-tsdocs generates descriptions

mint-tsdocs extracts frontmatter descriptions from your TSDoc comments:

/**
 * Centralized cache coordinator for TypeScript analysis and API resolution.
 *
 * @remarks
 * Provides production and development presets with configurable cache sizes
 * and optional statistics tracking for performance monitoring.
 */
export class CacheManager {
  // ...
}

Generates:

---
title: "CacheManager"
description: "Centralized cache coordinator for TypeScript analysis and API resolution"
---

This description appears in:

llms.txt (for AI tool indexing)
Page frontmatter (for SEO and navigation)
API reference page (below the title)

Writing effective descriptions

Classes

Focus on what it does and key features:

/**
 * Main orchestrator for documentation generation.
 *
 * @remarks
 * Coordinates template rendering, MDX generation, and navigation updates.
 * Processes API models from API Extractor into Mintlify-compatible documentation.
 */
export class MarkdownDocumenter {
  // ...
}

✅ Good: Explains role, lists main responsibilities, mentions output format

/**
 * Manages markdown documentation.
 */
export class MarkdownDocumenter {
  // ...
}

❌ Too vague: Doesn’t explain what “manages” means or what it produces

Interfaces

Describe what data it represents and where it’s used:

/**
 * Template variables available in Liquid templates.
 *
 * @remarks
 * Provides semantic variables (properties, methods, parameters) and metadata
 * (apiItem, page, navigation) for rendering API documentation pages.
 */
export interface ITemplateData {
  // ...
}

✅ Good: Explains purpose, lists variable categories, mentions usage context

/**
 * Template data structure.
 */
export interface ITemplateData {
  // ...
}

❌ Too generic: Doesn’t explain what template data this represents

Functions

Explain what it does and key parameters/behavior:

/**
 * Generates API Extractor configuration from mint-tsdocs config.
 *
 * @remarks
 * Auto-detects entry points, creates .tsdocs cache directory, and generates
 * api-extractor.json with proper paths for doc model output.
 */
export function generateApiExtractorConfig(config: ResolvedConfig): void {
  // ...
}

✅ Good: Clear action, mentions auto-detection, explains what it creates

Enums/Types

Describe possible values and use case:

/**
 * Error codes for documentation generation failures.
 *
 * @remarks
 * Includes validation errors (INVALID_CONFIG), file system errors (FILE_NOT_FOUND),
 * template errors (TEMPLATE_ERROR), and API extraction errors (EXTRACTION_FAILED).
 */
export enum ErrorCode {
  // ...
}

✅ Good: Explains purpose, groups error categories, provides examples

Best practices

Length

Aim for 10-25 words in the first line (what becomes the description):

/**
 * Centralized cache coordinator for TypeScript analysis and API resolution.
 * ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 * Perfect length - descriptive but scannable
 *
 * @remarks
 * Additional details go here...
 */

Specificity

Use concrete terms over vague ones: ✅ “Converts API models to template data for Liquid rendering” ❌ “Handles template data conversion” ✅ “LRU cache for type structure analysis with configurable size limits” ❌ “Cache implementation”

Keywords

Include TypeScript/API terminology that users might search for:

/**
 * Liquid template engine for rendering API documentation.
 *           ^^^^^^^ ^^^^^^         ^^^^^^^^^ ^^^
 * Keywords: liquid, template, engine, rendering, API, documentation
 */

Consistency

Use similar patterns for similar items:

/**
 * Centralized cache coordinator for TypeScript analysis and API resolution.
 */
export class CacheManager { }

/**
 * LRU cache for type structure analysis with eviction and statistics.
 */
export class TypeAnalysisCache { }

/**
 * LRU cache for API reference resolution with cached decorator support.
 */
export class ApiResolutionCache { }

All cache classes follow the pattern: “[Type] cache for [purpose] with [features]“

Checking your descriptions

After generating docs, review your llms.txt:

Generate and deploy

bun run mint-tsdocs generate
# Deploy to Mintlify

Access llms.txt

Visit https://your-docs-url/llms.txt to see the generated file.

Review API reference section

## API Reference

- [CacheManager](reference/mint-tsdocs.cachemanager): Centralized cache coordinator...
- [MarkdownDocumenter](reference/mint-tsdocs.markdowndocumenter): Main orchestrator...
- [ITemplateData](reference/mint-tsdocs.itemplatedata): Template variables available...

Check that descriptions are:

✅ Descriptive (not just repeating the class name)
✅ Scannable (10-25 words)
✅ Accurate (match what the class actually does)

Impact on AI tools

Before optimization:

User: "What caching options are available?"
AI: "I see CacheManager, TypeAnalysisCache, and ApiResolutionCache classes"
     [Can't explain what each does without reading full pages]

After optimization:

User: "What caching options are available?"
AI: "There are three caching components:
     - CacheManager: Centralized coordinator with presets
     - TypeAnalysisCache: For type structure analysis
     - ApiResolutionCache: For API reference resolution
     Each supports LRU eviction and optional statistics."

The AI can answer from llms.txt alone, without fetching all pages.

Updating descriptions

To improve a description:

Edit TSDoc in source code:

/**
 * [New description here]
 *
 * @remarks
 * Additional details...
 */

Regenerate documentation:

bun run build  # Build your TypeScript
bun run mint-tsdocs generate

Verify in llms.txt after deploying

Focus on your most commonly used classes first. Optimize CacheManager, MarkdownDocumenter, ITemplateData, etc. before less common utilities.

Learn more

Mintlify llms.txt Docs

Full llms.txt specification and features

TSDoc Reference

Supported TSDoc tags and syntax

MCP Integration

Real-time search alternative to llms.txt

Config Reference

Configure frontmatter generation

Get Started

Usage

Components & Snippets

AI Tools

llms.txt Optimization

Why llms.txt matters for mint-tsdocs

How mint-tsdocs generates descriptions

Writing effective descriptions

Classes

Interfaces

Functions

Enums/Types

Best practices

Length

Specificity

Keywords

Consistency

Checking your descriptions

Impact on AI tools

Updating descriptions

Learn more

Mintlify llms.txt Docs

TSDoc Reference

MCP Integration

Config Reference

​Why llms.txt matters for mint-tsdocs

​How mint-tsdocs generates descriptions

​Writing effective descriptions

​Classes

​Interfaces

​Functions

​Enums/Types

​Best practices

​Length

​Specificity

​Keywords

​Consistency

​Checking your descriptions

​Impact on AI tools

​Updating descriptions

​Learn more

Mintlify llms.txt Docs

TSDoc Reference

MCP Integration

Config Reference

Why llms.txt matters for mint-tsdocs

How mint-tsdocs generates descriptions

Writing effective descriptions

Classes

Interfaces

Functions

Enums/Types

Best practices

Length

Specificity

Keywords

Consistency

Checking your descriptions

Impact on AI tools

Updating descriptions

Learn more