How to Get Started with KoboldAI
Updated
To get started with KoboldAI, decide first whether you want to run a local LLM server or use a hosted web interface. The official site presents four foundations: KoboldCPP (local LLM API server), KoboldAI Lite (lightweight interface for API endpoints), KoboldAI.net (KoboldAI Lite delivered as a free web service), and the Discord Server for community support. The site's Start section is the intended entry point for everything you need to begin.
Choose Your Starting Point
| Option | What it is | Best if you want |
|---|---|---|
| KoboldCPP | Local LLM API server for driving your backend | To run models on your own hardware |
| KoboldAI Lite | Lightweight, user-friendly interface for accessing AI API endpoints | A front-end that connects to any compatible endpoint |
| KoboldAI.net | KoboldAI Lite delivered as a web service, free, with the same flexibilities as running it locally | To try it without installing anything |
If you are unsure, start with KoboldAI.net to test the interface, then move to KoboldCPP when you want local control.
Step 1: Read the Overview
The Overview section states KoboldAI's purpose plainly: it "delivers a combination of four solid foundations for your local AI needs." Read this first so you understand which component does what before downloading anything.
Step 2: Follow the Start Section
The site directs beginners to its Start section: "Everything you want to know for your start with KoboldAI's solutions you find on this website." This is the recommended path from overview to actual use.
Step 3: Download KoboldCPP (for local use)
If you are running locally, go to Download KoboldCPP, which links to the official downloads page for Windows, Linux, and Mac. The site also lists a KoboldAI Quickstart guide under Guides, alongside Supported Models, AI Safety, KoboldCpp API, and KoboldCpp Administration.
Additional KoboldCPP guides on the site
- KoboldCpp Quicklaunch
- KoboldCpp Tokens
- KoboldCpp on Google Colab
- KoboldCpp on Runpod
- Networking
- Audio
- Hardware options compatible with Kobold
- Horde Worker
- Image Generation
- Loaded Files
Step 4: Join the Community
The site recommends joining the Discord Server for "support, discussions, and sharing." This is the fastest route when a guide does not cover your specific hardware or setup.
Common Starting Points to Check
- Hardware options compatible with Kobold — confirm your machine is supported before downloading.
- Supported Models — check which models work with your chosen setup.
- AI Safety — review before deploying anything publicly.
- KoboldCpp on Google Colab / Runpod — use these if your local hardware is insufficient.
The site does not list pricing for any component; KoboldAI.net is described as a free web service, while local options depend on your own hardware and model choices.