| Flagship Autonomous Business Starter | - Indie builders who want to launch a niche directory or review site that an AI agent keeps up to date on a schedule
- People already using Claude Code, Cursor, Codex or a similar coding agent who want ready-made operating rules and routines for it
- Anyone who wants to start on free tiers and only pay fees when something sells
| - Astro 5 + Tailwind 4 static directory site: category and tool pages, comparisons, guides, a "new this week" feed, newsletter, advertise and submit pages, sitemap and JSON-LD
- Sample data to start from: 3 categories, 15 tools and 6 guides, all in plain JSON/JS files
- Scout script: collects new candidate tools from Hacker News, GitHub and Product Hunt into a review queue (Python standard library, no API keys)
- Evidence script: scores each tool on a public rubric (free tier, pricing, docs, platforms, open source, recent updates). The free-tier tick comes from the kit's own data file, not a live check. A tool with no public repository scores zero on the open-source part, and that zero lowers the score. A criterion the script cannot check, such as a page that blocks it, is left out of the score
- Routine prompts: daily new tools, weekly newsletter draft, weekly social posts, monthly health check
- Approval-gate rules and AGENTS.md / CLAUDE.md for your coding agent
- Stripe Payment Links wiring (fast-track review, newsletter sponsor, featured listing) + newsletter signup embed
- Step-by-step setup guide (zero to live on free tiers) and runbook
| US$49 |
| MCP Server Pack | - Developers using Claude Desktop, Claude Code or Cursor who want their assistant to work with notes, web pages, GitHub issues, a SQLite database, news feeds or a local model
- Anyone learning to write MCP servers who wants small, readable, working examples on the official TypeScript SDK
- People who want safe defaults (sandboxed, read-only unless enabled) before handing tools to an AI assistant
| - filesystem-notes: list, read, write and search Markdown notes, sandboxed to one folder
- web-fetch: clean article text from any page (Mozilla Readability), with private/LAN URLs blocked
- github-issues: list, search and read issues with comments; create and comment only when you opt in
- sqlite-query: tables, parameterised read-only queries, opt-in writes (WebAssembly, so no native build on Windows)
- rss-news: merge RSS and Atom feeds newest-first, with a keyword filter
- ollama-bridge: chat, model list and embeddings from your local Ollama
- Per-server READMEs + Claude Desktop, Cursor and Claude Code configs
- End-to-end smoke test (22 checks) and step-by-step notes for building your own server from any of the six
| US$29 |
| Local-LLM App Starter | - Anyone who wants a ChatGPT-style assistant on their own computer, with chats and documents that stay on the machine
- Developers who want a small, readable base for a local AI app: a plain Node.js backend with no runtime dependencies and a vanilla JavaScript UI
- People with an NVIDIA GPU, an Apple Silicon Mac or a CPU-only laptop who aren't sure which model their hardware can run
| - Streaming chat UI with model picker, stop and clear (Ollama); your model choice is remembered
- Ask-my-documents Q&A: add .txt, .md, .csv or code files, embedded locally, with answers that cite their sources
- Which-model-fits-my-GPU tool (detects NVIDIA VRAM or Apple unified memory) + VRAM guide covering 15 models from 1B to 70B
- One-command setup: setup.ps1 (Windows) and setup.sh (Linux/macOS)
- Smoke tests and a mock mode (no model needed)
- Notes for extending it: PDF support, and swapping the JSON vector store for SQLite or LanceDB
| US$39 |
| Agent Safety & Hooks Pack | - Developers who let Claude Code or Cursor run commands and edit files in real projects
- Solo builders and small teams who want a confirmation step before an agent pushes, deploys or publishes
- Anyone who wants commits blocked when they contain secrets, Claude Code blocked from writing secrets into files, and both tools blocked from reading .env and key files
| - Command guard: blocks rm -rf /, force-push to main, git reset --hard, curl | sh, DROP DATABASE and more
- Approval gates: asks before git push, deploys, publishing, data-sending curl, migrations, sudo and deletes
- Secret guard: on Claude Code, blocks API keys, tokens and private keys in agent edits, and blocks reads and writes of .env and key files. On Cursor, the hook blocks reads of .env and key files only; it does not block writing secrets into ordinary files. A Cursor rule asks the agent not to hard-code secrets
- Cost/token budget guard with soft (ask) and hard (stop) limits
- Redacted JSONL audit log of every tool call
- Git pre-commit secret scanner + one-command installer
- Claude Code settings.json + CLAUDE.md, and Cursor hooks.json + 4 .mdc rules
- Config file for your own patterns, and a 57-check test suite
| US$19 |
| Agent-Ready Rules Bundle | - Developers using Cursor, Claude Code, Codex, Copilot, Windsurf, Aider or another agent that reads AGENTS.md
- People who want their agent to follow the stack's conventions and run lint, type checks, tests and build before saying it's done
- Anyone working across several stacks who wants the same quality and safety rules in every repo
| - 8 stacks: Next.js, Astro, FastAPI, Node API, Expo, Go, Rust, Tauri
- AGENTS.md + CLAUDE.md + Cursor rules for each stack
- Each stack folder: AGENTS.md, a CLAUDE.md that imports it, a Cursor rule that attaches to that stack's files, and an always-on quality and safety rule
- Safety guardrails and per-stack check commands
- Free lite edition (Next.js + FastAPI): direct download at /free/
| US$15 |
| Pro Prompt Pack | - People who use ChatGPT, Claude, Cursor or local models every day and want reusable prompts instead of starting from a blank box
- Claude Code and Cursor users who want ready-made slash commands for reviews, plans, commits and root-cause hunts
- Local-model users who need prompts short and strict enough for small 1B-9B models
| - 209 prompts as Markdown (by category) + one JSON file
- Categories: vibe coding (32), local LLM (25), indie marketing (25), agents (23), debugging (23), writing (22), code review (21), research (21), gaming & content (17)
- 10 chained agent workflows: feature build, bug hunt, code quality, launch week, pre-launch, local RAG, safe agent setup, research report, content engine and game content
- 12 slash commands for Claude Code (.claude/commands) and Cursor (.cursor/commands): /plan, /review, /rootcause, /tests, /security, /commit, /prdesc, /docs, /explain, /cleanup, /shipcheck, /agentsmd
- Model notes for each prompt, including prompts tuned for small local models
- Free preview: a selection of the prompts at prompts.lilaforge.com
| US$12 |