Your classroom tools. Your computer. Your judgment.
Local School helps you draft lessons, classroom materials, and feedback using local AI on your computer. Choose a tool, give it classroom context, and review the result before using it.
This guide covers the Apple Silicon and Intel Mac packages (both version 0.8.0, build 9) and a supplied Windows x64 0.8.0 beta installer. Windows privacy behavior is not source-verified because the supplied ZIP contains no application source. Start with fictional examples, and follow your school's requirements before using student information.
1. Choose your desktop version
The individual desktop app has a $20 USD one-time base price plus applicable tax. Mac packages and the signed Windows beta package are supplied. Check the purchase flow for current Stripe checkout availability; after successful payment is verified, choose the package for your computer:
The purchase terms are for one teacher on up to two computers for home and work use, across the supported builds. Monthly updates are included for one year from successful payment and are provided manually by email; the license to the purchased version does not expire after that year. There are no automatic renewals or recurring Local School fees; later updates would require separate terms. Request a no-questions-asked refund within seven days of purchase by emailing team@localschool.ai; refunds are handled manually by email. No justification or additional usage condition is required, and there is no fee to request a refund. Read the purchase terms and purchase privacy notice.
- Apple Silicon Mac: On the Apple menu, choose About This Mac. If it shows Chip (for example, an Apple M-series chip), choose the Apple Silicon ZIP. Requires macOS 13 or later.
- Intel Mac: In About This Mac, look for Processor showing Intel. Choose the separate Intel ZIP. Requires macOS 14 or later. The app runs models using CPU on Intel; begin with Quick start. Intel performance has not been validated.
- Windows PC: According to the supplied package README, this is a Windows 11 x64 beta. It requires an x64 processor, at least 8 GB RAM for Quick start, and at least 2 GB free disk space for the app, starter model, and download overhead. CPU inference is supported; no dedicated graphics card is required. Quick start downloads about 1.1 GB. Larger models need more memory and disk space; performance depends on the computer.
You'll need internet to download and install the app and to download an AI model. In the inspected Mac source, generation can run offline after a model is downloaded and loaded. The Windows README says its model downloads need internet and generation runs locally, but its application source was not included for verification. No separate AI app is required; the README reports no Windows account requirement.
2. Install Local School
The Mac packages and Windows x64 beta ZIP are supplied. After Stripe confirms a payment and the server verifies it, use the protected download page to choose the matching package:
- Use About This Mac to select the Apple Silicon or Intel ZIP that matches your Mac and confirm it meets the macOS minimum above.
- Download the matching ZIP and double-click it in Finder to extract it.
- Move Local School.app into your Applications folder.
- Open Local School from Applications.
The publisher reports the Mac packages are notarized. This is not Apple endorsement or a privacy certification, and this guide does not independently verify Gatekeeper acceptance or provide an independent Mac runtime assessment. If macOS blocks the app or says it cannot be checked, stop and contact support. Do not bypass Gatekeeper, use an override, or disable your Mac's security protections.
For the supplied Windows beta, follow the steps in its README:
Windows 0.8.0 beta signing: Local School for Windows is digitally signed and offered by Adaptable Learning. Its code-signing certificate allows Windows to verify the publisher and detect changes to the signed installer. Both the application and installer have owner-verified digital signatures with Microsoft timestamps; the signing service is Microsoft Azure Artifact Signing. This is owner-supplied verification, not an independent cryptographic or runtime audit by this website. Code signing verifies publisher identity and signed-installer integrity only; it is not a Microsoft license, endorsement, app-quality certification, independent application security audit, or platform-compatibility guarantee. Windows or SmartScreen may still warn.
- Right-click the downloaded ZIP and choose Extract All.
- Open the extracted folder and run Local-School-0.8.0-Windows-x64-Setup.exe.
- Follow the installer steps, then open Local School from the desktop shortcut or Start menu.
- In Model library, download and load Quick start. The README says this first model download requires an internet connection and generation runs locally; this is not source-verified.
If Windows or SmartScreen warns, or your school's security policy blocks installation, stop and ask your school IT team or Local School support. Do not select “Run anyway,” bypass or disable SmartScreen, or weaken other protections. Stop if the publisher is missing or differs from the expected name, or Windows reports an invalid signature.
The publisher-provided SHA256SUMS.txt in the ZIP lists this installer hash: 6c987ed3b1b0b9ba071683c7da720be79bf268181539535314fd19df0b0acc7b. This supplied checksum is not an independent authenticity check.
The Windows ZIP README says the installer reuses the existing local data folder to preserve models and curriculum, and uninstalling the app preserves that data. Keep your own copies of exports. This data-handling statement is publisher-reported; Windows application source was not included for inspection.
3. Download and load a model
A model is the AI software that creates your drafts. It runs inside Local School. The app includes these model choices; suggested memory is guidance, not a guarantee of performance:
On Intel Mac, model generation uses CPU. The app advises starting with Quick start because CPU generation may take longer; Intel speed has not been validated. The Teaching pilot is described as an Apple Silicon option.
Open Model library and choose an option:
| Model | Download | Suggested memory | Start here if… |
|---|---|---|---|
| Quick start · Qwen 2.5 1.5B | About 1.1 GB | 8 GB or more | You want the smaller download to try the tools. Not validated for grading. |
| Larger model · Qwen 2.5 7B | About 4.7 GB | 16 GB or more | You have more memory and can allow extra time. Not validated for grading. |
| Teaching pilot · gpt-oss 20B | About 12.1 GB | 24 GB or more | You are testing the current gpt-oss option on Apple Silicon. Windows performance has not been validated; feedback and quizzes still require careful teacher review. |
Choose Download model ↓ and keep the app open while it downloads. The app checks the downloaded model files before they can be loaded. When the download is verified, choose Load model and wait for LOADED ON THIS COMPUTER.
Downloading and loading are separate steps. On later visits, load the model already on your computer. Once downloaded and loaded, you can generate drafts offline. Performance depends on your computer, available memory, model, and request length. No model has been validated for dependable grading.
4. Create your first classroom draft
- Open Teacher toolkit.
- Choose Lesson planner.
- Select Try an example, or complete the required fields with your own lesson context.
- Select Create draft locally.
- Read the draft and edit it in Make it yours. Check facts, timing, directions, and fit for your students.
- Check I have reviewed this draft for accuracy and classroom use.
- Choose Word document (.docx) and select Export reviewed draft ↗. Pick a save location.
Some tools display their output only after the whole draft is ready. You can use Stop to cancel generation.
Export before closing. Inputs and drafts stay available while you switch tools, but this version does not save projects. Changing a tool's inputs clears its previous draft; export anything you want to keep first.
5. Add your curriculum as reference material
Use a course outline, unit guide, or classroom reading to give Local School more relevant context.
- Open Curriculum library.
- Select Add documents from this computer.
- Choose a PDF, Word (.docx), Markdown (.md), or plain-text (.txt) document.
- Select Preview text and check that the imported text is complete.
- Check Use for responses for each document you want to include. New documents are not selected automatically.
- Return to a teacher tool or Ask Local School. Name the topic clearly in your request.
Local School searches selected documents for matching passages and includes relevant excerpts when they fit. Check the source excerpts listed with the response. Search can miss information, and a listed reference does not prove every generated claim is correct.
This supplies reference material; it does not retrain the model. Scanned pages, images, and diagrams are not read. Use a document with selectable text, and check the preview. Curriculum text and selections stay saved locally between sessions.
6. Explore the assessment tools
Paper feedback
Open Teacher toolkit → Assess → Paper feedback. Select Load sample paper for a fictional demonstration, or enter an assignment, rubric, and paper. You can paste a paper or use Import .txt; this tool does not currently import Word or PDF directly.
Select Draft feedback. Review the evidence, scores, and wording, then edit the feedback. Check the review box before exporting. You make the final assessment decisions.
Defend Your Learning
Open Teacher toolkit → Assess → Defend Your Learning. Paste a paper or select Import student paper to use a PDF, Word (.docx), Markdown (.md), or text (.txt) file. Check imported text before generating.
The tool creates five comprehension questions and a separate teacher guide grounded in the paper. It uses only that paper, even when curriculum documents are selected elsewhere.
Review the questions and teacher guide before use. Exports include the guide, so remove teacher answers from student copies. Use the questions to discuss understanding—with appropriate thinking time and support. They cannot establish whether a student used AI or who wrote the paper.
7. Choose an export format
| Format | Useful for |
|---|---|
| Word (.docx) | Editing and adapting your draft in a word processor. |
| PDF (.pdf) | Printing or sharing a fixed-layout copy. |
| Plain text (.txt) | Copying into another classroom tool. |
| Markdown (.md) | Keeping text with Markdown formatting markers. |
| Web document (.html) | Opening the exported document in a browser. |
All 17 teacher tools support these formats. Exports are text documents; slide outlines do not become PowerPoint files, and text-based tables do not become formatted Word tables. Review the exported file before sharing it.
What stays on your device?
The privacy and storage notes below are based on source bundled in both supplied Mac 0.8.0 packages. The supplied Windows ZIP contains an installer, README, and checksum, but no application source; its privacy behavior has not been independently inspected. For more detail, see Desktop privacy.
- AI generation: Runs inside Local School on your computer. The inspected Mac app source does not send papers, prompts, curriculum text, or generated drafts to a cloud AI service.
- Model downloads: Connect to Hugging Face and its download infrastructure for public model files. The downloads do not include your teaching content.
- Drafts, papers, and chat: Held in the app workspace and cleared when you close the workspace. Export work you want to keep. Clearing is not a secure-erasure guarantee.
- Curriculum library: Extracted text, filenames, and selections are saved locally between sessions. The app does not encrypt this library. Removing a document moves its saved record to your Mac’s Trash; original files and existing exports remain unchanged.
- Models and exports: Downloaded models stay on your computer until you delete them. Exports are saved wherever you choose. A folder synced with iCloud, Google Drive, Dropbox, or another service may upload an export.
Open Privacy & storage in the app for more detail. Local processing does not replace your school's approval process or your device's security and backup settings.
Quick fixes
| If you see… | Try this |
|---|---|
| No model loaded, or generation is unavailable | Open Model library and choose Load model for a downloaded model. |
| A slow response | Allow more time, close other memory-heavy apps, or try Quick start with a shorter request. Intel Mac uses CPU and may take longer. |
| An import with missing text | Check Preview text. Use selectable text rather than scanned pages or images. |
| A request that's too long | Shorten it intentionally and check what you removed. The app does not silently shorten oversized input. |
| An unavailable export button | Wait for generation to finish, review the draft, and check the review box. Editing the draft requires review again. |
| An incomplete or failed response | Do not treat it as finished. Try a shorter request or generate again and review the new result. |
| A chat reply missing earlier context | Include the necessary context in your next message. Each request is independent in this version. |
Need help?
For help, email team@localschool.ai. Support is email only; we do not offer phone or live chat support. We’re a small team, so responses may take time. We appreciate your patience.
Include your platform (Apple Silicon Mac, Intel Mac, or Windows PC), tool name, app version, model, memory, and the error wording. Use a fictional example to reproduce the issue. Do not send student papers, names, grades, or other student information; remove it from screenshots too.
Need help? team@localschool.ai · Email only. We're a small team; responses may take time.