Skip to main content

PitchDocs

Perfect repo docs

Documentation skills and templates for AI coding assistants.

Scans your codebase and generates professional docs — README, CHANGELOG, CONTRIBUTING, llms.txt, and more. Every document answers the four questions readers actually ask.

Releasedv1.7.0
Get Started View on GitHub

What it does

Slash commands that turn your codebase into docs people actually read

README generation

Scan your codebase and produce a README that answers: what does this do, does it solve my problem, can I use it, where do I learn more?

A README that sells your project, not just describes it

Documentation audit

Audit 20+ file types across 3 tiers (Essential, Professional, Mature) plus GitHub metadata, visual assets, and package registry config.

A complete docs audit with specific gaps identified

Changelog automation

Transform conventional commits into user-benefit language. Readers see what changed for them, not your commit log.

Changelogs written in benefit language users actually read

llms.txt generation

Generate llms.txt and llms-full.txt following the llmstxt.org spec — make your docs discoverable by AI coding assistants.

llms.txt files that make your project AI-discoverable

What PitchDocs includes

  • Evidence-based feature extraction

    Scans 10 signal categories and surfaces selling points with file-level evidence.

  • 4-question framework

    Every doc answers: what does this do, does it solve my problem, can I use it, where do I learn more?

  • 20+ file documentation audit

    Catch missing docs, broken metadata, and invisible images before you ship.

  • 10 slash commands

    Generate any doc type in under a minute, from README to llms.txt.

  • Ready-to-use templates

    CONTRIBUTING, SECURITY, SUPPORT, issue/PR templates, release config — one plugin replaces 10+ files.

  • llms.txt generation

    AI-readable content indices so coding assistants and search engines surface your docs.

  • GitHub community templates

    Generate CONTRIBUTING, SECURITY, SUPPORT, and issue/PR templates — one command sets up all your community health files.

  • npm and PyPI compatibility

    Audit registry metadata and cross-renderer README compatibility.

  • Progressive disclosure

    Non-technical language first, technical depth deeper, with automatic cross-linking.

  • Upstream spec drift detection

    GitHub Action checks monthly that your docs follow the latest spec versions.

How it works

PitchDocs runs entirely inside your AI coding assistant — no external dependencies, no API calls

100% local
Evidence-based
4-question framework
Fully open source
What exactly does it access?

What it accesses

  • Your local codebase (to extract features and capabilities)
  • Existing documentation files in your repo

Where it stores

  • Generated files are written to your repo directory

Delete generated files or git restore to undo

Network calls

  • None — pure markdown plugin with no external requests

Nothing to disable

Get started in 60 seconds

Install via the LBA plugin marketplace — same steps work in both tools. OpenCode reads .claude/skills/ natively.
Using Codex CLI, Cursor, Gemini CLI, Aider, or Goose? See the setup guide.

1

Add the LBA plugin marketplace

/plugin marketplace add littlebearapps/lba-plugins

One-time setup — works in both Claude Code and OpenCode

2

Install PitchDocs

/plugin install pitchdocs@lba-plugins

Pure markdown plugin, no build step

3

Generate your first README

/readme

Scans your codebase and generates a marketing-quality README

🔧
Setup for Other AI Tools
📖
Full Documentation

Ready to go? Try these commands

Copy one of these into Claude Code to see it in action

"/readme"

Scans your codebase and generates a marketing-quality README

"/features"

Extracts features and benefits from 10 signal categories

"/docs-audit"

Audits 20+ file types across 3 maturity tiers

"/changelog"

Transforms conventional commits into user-benefit language

"/llms-txt"

Generates llms.txt and llms-full.txt for AI discoverability

"/roadmap"

Creates a public roadmap from GitHub milestones and issues

"/ai-context"

Generates AI context files — CLAUDE.md, AGENTS.md, .cursorrules

"/docs-verify"

Checks documentation quality, links, and freshness

"/launch"

Creates launch checklists with community and registry tasks

"/user-guide"

Generates end-user documentation from your codebase

"I kept writing the same boilerplate documentation for every new project — README, contributing guide, changelog, security policy. And the result was always mediocre."

N
Nathan

Where to find PitchDocs

IssuesReport bugs
DocumentationGuides & API reference
DiscussionsCommunity & questions
View on GitHub

Questions about PitchDocs

What is PitchDocs?
PitchDocs is a documentation plugin that generates and audits public-facing repository documentation. It produces README files, changelogs, roadmaps, contributing guides, and llms.txt files designed to attract users and contributors. Works with Claude Code, OpenCode, Cursor, Codex CLI, Gemini CLI, Aider, and Goose.
How is PitchDocs different from other documentation generators?
Most generators produce technically-accurate but dry output. PitchDocs applies the 4-question framework — every document answers: what does this do, does it solve my problem, can I use it, where do I learn more? The result is docs that sell as well as they inform.
Does PitchDocs work with any programming language?
Yes. PitchDocs scans your codebase for 10 signal categories (CLI commands, public API, configuration, integrations, performance, security, TypeScript/DX, testing, middleware/plugins, and documentation) regardless of language.
Does PitchDocs need an internet connection?
No. It's a pure markdown plugin with no runtime code and no external API calls. Everything runs locally inside your AI coding assistant.
What is the 4-question framework?
Based on the Daytona/Banesullivan approach used by high-star repos like PostHog, Fiber, and LobeChat. Every document must answer four reader questions: What does this do? Does it solve my problem? Can I use it? Where do I learn more?
Can I use PitchDocs with other AI coding assistants?
Yes. PitchDocs works with Claude Code (primary), OpenCode, Cursor, Codex CLI, Gemini CLI, Aider, and Goose. Install via the LBA plugin marketplace — the skills and templates are compatible across tools.
What's included in a documentation audit?
PitchDocs audits 20+ file types across 3 tiers (Essential: README, LICENCE, .gitignore; Professional: CONTRIBUTING, CHANGELOG, SECURITY; Mature: governance docs, release config). It also checks GitHub repository metadata, package registry configuration (npm/PyPI), and visual assets.
Does PitchDocs overwrite my existing docs?
Generated files are written to your repo directory. Use git diff to review changes before committing. You can always git restore to undo.

What changes

  • • Instead of writing README boilerplate from scratch, PitchDocs scans your codebase and generates documentation that answers the questions readers actually ask.
  • • Instead of manually checking which docs you're missing, run a 20-file audit that catches gaps across three maturity tiers.
  • • Instead of copying your git log into a changelog, PitchDocs transforms conventional commits into user-focused release notes.
  • • Instead of guessing how to make your project AI-discoverable, generate llms.txt and AGENTS.md following the official specs.

Roadmap

Coming Soon

  • Exploring: Figma/design asset documentation support
  • Exploring: Multi-language README generation

Changelog

v1.7.0
github-actions[bot]github-actions[bot]
  • •add /doc-refresh command for version-bump documentation updates (58af69b)
  • •add AGENTS.md spec to upstream version tracking with v1.1 feature monitoring (c8c9fa3)
  • •add cross-platform AI tool setup instructions (3746e05)
  • •add cross-platform AI tool setup instructions and AGENTS.md (b0253df)
See release details on GitHub
v1.4.1
github-actions[bot]github-actions[bot]
  • •add content filter mitigations and visual formatting guidance (0e89233)
  • •add missing colour to version badge in README (92a679c)
  • •use query-param badge URL so release-please preserves colour (f39c1f0)
See release details on GitHub
View all releases on GitHub →

Also by Little Bear Apps

Untether Run Claude Code, OpenCode, or Codex CLI from Telegram. Voice notes, multi-project routing, and persistent sessions. Outlook MCP 20 consolidated tools for email, calendar, contacts, and more — all through natural language.
← Back to All Tools