The pitch in 90 seconds
Most AI starter kits hardcode Claude or GPT and call it a day. That works until the API bill arrives. Or until you want to run on infrastructure you control. Or until you want to ship a free tier without paying per call.
The 20 wrappers launching today fix that. Each is a small working app — paste text, get AI output — and each ships with two paths:
- Commercial API path (5-minute setup, costs pennies per call)
- Free open-source model path (15-minute setup, $0 per call, runs on your laptop)
Same code. Same prompts. The only difference is the transport — Claude/GPT for the commercial path, Llama 3.1 / Whisper.cpp / rembg / Stable Diffusion for the open-source path. A layman-language guide in every wrapper walks through the swap. No ML jargon. No assumed expertise.
The catalog at launch:
- 5 free wrappers — email-gated downloads, live demos in your browser at clskillshub.com/wrapper/<slug>
- 15 paid wrappers at $20 each
- All-access bundle at $99 for all 15 paid (saves $201)
This post: what's in the catalog, how the open-source paths work, who this is for, and where to start.
The 5 free wrappers (live now)
Every free wrapper has a hosted demo you can try without entering an email. The download (working code + integration guide + sample data) requires an email — that's the only gate.
1. Text Summarizer
Paste any article, blog post, transcript, or meeting notes. Get a 3-line TL;DR plus 5 bullet takeaways. Claude path uses Claude Haiku 4.5; open-source path uses Llama 3.1 8B via Ollama. Stack: Next.js. Setup: 5 minutes.
2. Email Reply Drafter
Paste an inbound email. Get three reply variants: direct, warm, decline. Useful for clearing inbox debt — pick the variant that fits your relationship to the sender. Free path: Llama 3.1 via Ollama.
3. Meeting Notes Summarizer
Paste a raw transcript. Get decisions, action items with owners, open questions, key topics. Handles speaker labels (Samarth: ...) if your transcript has them; works without if it doesn't. Free path: Llama 3.1.
4. Cover Letter Generator
Paste a JD and 3-5 resume highlights. Get a 3-paragraph cover letter under 350 words. Tuned to avoid the AI tells (no 'I am thrilled', no 'passionate', no 'synergy'). Free path: Llama 3.1.
5. Job Description Writer
Enter a role title and 3 must-haves. Get a full JD with seniority calibration, inclusive-language check, and an honest day-to-day section. Free path: Llama 3.1.
All 5 free wrappers use the same Next.js skeleton with different prompts. Once you've installed one, the others install in seconds.
The 15 paid wrappers ($20 each, $99 bundle)
These ship over the next four weeks per the public roadmap. Buyers today get every wrapper as it lands via the same unlock link — no re-buy.
Text-heavy ($20 each — Node.js)
- PDF Q&A Bot — RAG over any PDF with citations. Free path: Nomic Embed + Llama 3.1.
- Customer Support Bot — RAG over your docs, deploy as a widget. Free path: bge-small + Mistral 7B.
- Cold Email Personalizer — LinkedIn URL → personalized email. Free path: Llama 3.1 70B via Groq free tier.
- SEO Article Generator — keyword → 1,000-word article with H2/H3 structure. Free path: Llama 3.1 70B.
- Code Documenter — paste a function, get docstring + comments. Free path: DeepSeek Coder via Ollama.
- Resume Rewriter — resume + target role → ATS-optimized rewrite.
- Product Description Generator — e-commerce product details → full listing copy.
- Tweet Thread Generator — topic → 8-tweet thread with hooks.
- Quiz Generator — text → 10 questions with answer key.
Audio ($20 each — Python)
- Voice Note Transcriber — audio in, cleaned transcript with speaker labels. Free path: whisper.cpp / faster-whisper, runs on CPU.
- YouTube Video Summarizer — URL → summary + key points + timestamps. Free path: yt-dlp + whisper.cpp + Llama 3.1.
OCR + Image ($20 each — Python)
- OCR + Cleanup — image with text → structured clean text. Free path: Tesseract or PaddleOCR.
- Background Image Remover — image in, transparent PNG out. Free path: rembg (ONNX), runs on CPU.
- AI Sticker / Avatar Generator — text prompt → stylized image. Free path: Flux.1-schnell or SDXL Lightning (GPU helps).
- Receipt / Invoice Data Extractor — image → structured JSON. Free path: DONUT or LLaVA + Tesseract.
The $99 all-access bundle unlocks all 15 paid wrappers in one ZIP plus every future wrapper at no extra cost. Anyone who plans to use 5+ of the paid wrappers should buy the bundle — saves $201 vs paying individually.
How the open-source paths actually work
This is the part most AI wrapper kits skip. Every wrapper in this catalog has a guide.md that walks through the specific open-source swap for that wrapper's task. The guides assume zero ML background. Here's the general pattern.
For text wrappers (1-9 and 11-14): install Ollama, pull llama3.1:8b (4.7 GB, one-time download), set USE_OLLAMA=true in the wrapper's .env.local. Done. Output quality is roughly 90% of Claude's for these tasks; for high-stakes content (top-tier-firm cover letters, regulated copy), Claude still wins. For volume work (bulk JDs, daily email triage), Llama is more than sufficient and costs nothing.
For audio (15, 16): install whisper.cpp or faster-whisper. Both run on CPU; faster-whisper is ~3x faster but needs a slightly more complex install. Quality is essentially identical to OpenAI's Whisper API because under the hood it's running the same model architecture, just locally.
For OCR (17): install Tesseract via your OS package manager (brew install tesseract on Mac, apt install tesseract-ocr on Linux). For higher accuracy on tricky inputs (handwriting, low-quality scans), the wrapper also supports PaddleOCR — slightly more setup, materially better output.
For background removal (18): install rembg via pip. Runs entirely on CPU using ONNX models. Quality is competitive with Removal.ai or Cloudinary for most subjects; struggles slightly on hair and fur compared to dedicated paid services. The wrapper documents which rembg model to pick by subject type.
For image generation (19): this one is honest about the trade-off. Running Stable Diffusion XL or Flux.1-schnell locally needs a GPU. The wrapper documents how to run it on a $0.30/hr rented GPU (Vast.ai, RunPod, Modal), how to run it on Replicate ($0.005-0.02 per image), and how to run it on your own RTX 3090 if you have one. No live browser demo for this category — the wrapper ships with a Loom walkthrough instead.
For receipt extraction (20): the wrapper supports two paths. The fully-local path uses DONUT or LLaVA, both need GPU. The pragmatic path uses Tesseract for the OCR step and Claude for the structured-JSON extraction — Tesseract is free, Claude costs about $0.005 per receipt. The guide explains the trade-offs.
Who this is for
The wrappers are for developers shipping AI features who care about the marginal cost per call. That includes:
- Indie hackers building tools where the customer pays $5-50/month and the AI cost can't be 30% of revenue.
- Bootstrapping SaaS founders prototyping features they'd burn cash on with paid APIs alone.
- Internal-tools developers at companies where buying an API subscription requires a procurement cycle.
- Self-hosters and privacy-conscious builders who want the model running on their own machine.
- Anyone learning to deploy AI without committing to a vendor.
The wrappers are NOT for:
- Pure prompt-engineering buyers — for that, the $49 prompt packs (Legal, SAP, Recruiter, etc.) are a better fit. Those are 50 tested prompts per pack, not working apps.
- Enterprise teams with existing vendor commitments — if your company is on a $50K/year Azure Cognitive Services contract, swapping to local models is a procurement problem, not a code problem.
- Builders who only want managed SaaS UIs — these wrappers are starter code, not SaaS products. You run them yourself.
Where to start
If you've never run a local AI model: start with the Text Summarizer. It's free, the demo runs in your browser without an email, and the guide walks through installing Ollama for the first time. Once you have Ollama installed, every other Llama-based wrapper in the catalog installs in seconds.
If you're already running Ollama or local models: skip straight to whichever paid wrapper matches your use case. The Voice Note Transcriber is the highest-leverage one for most people — Whisper API costs add up fast and whisper.cpp is genuinely free.
If you want everything: the $99 all-access bundle is the math choice. 15 wrappers at $20 each = $300; bundle = $99. Plus every future wrapper free.
Why this exists
clskillshub.com has been shipping prompt-engineering packs for working professionals for two years. The packs work because the buyer already has the AI access — they just need better prompts.
The wrappers are the next step for buyers who don't just want better prompts but want working code. The open-source-path framing is the differentiator: most AI starter kits assume you'll pay for the API forever. These don't.
Three more things worth saying:
- The free wrappers fund the catalog. They feed the newsletter list which sustains the rest. If you want to support this work without buying, just download a free wrapper.
- The open-source guides will keep getting better. Every wrapper's
guide.mdis a v1.0; v1.1 ships within 30 days of launch with refinements based on what early buyers ask about. - Buyer feedback shapes the next 20. I have a list. You probably have wrappers I haven't thought of. Reply to your delivery email and your suggestion shapes the v1.1 expansion list.
Related reading
- The 7 Patterns Behind High-Performance Claude Prompts — the prompt framework every wrapper uses
- Harvey AI Alternative for Lawyers — the prompt pack series this product line extends
- Pricing page — every product on the site
- Wrappers catalog — all 20 wrappers
FAQ
Do the free wrappers really work without paying anything? Yes. The free wrappers download with working code, integration guide, and sample data. The hosted browser demo uses Claude Haiku 4.5 (which I pay for, rate-limited per IP), and the included code lets you swap to free Llama 3.1 in 15 minutes.
Why $20 for the paid ones and $99 for the bundle? $20 is the price point where the math always favors buying over building from scratch — even 1 hour of dev time is worth more. $99 anchors the bundle at "obvious buy" for anyone who'd use 5+ wrappers; saves them $201.
What if I want a wrapper not on the list? Reply to your purchase email (or to my newsletter address). Your suggestion shapes v1.1 expansion. The next 20 wrappers ship based on buyer demand, not my guess.
Refund policy? Digital product, all sales final. If something genuinely doesn't land for you, reply to your purchase email and I'll add you to the full Skills Library (lifetime access) as a goodwill gesture.
Lifetime updates? Yes. Same unlock link serves every future version.