NanoGPT + Open WebUI Setup: Self-Hosted AI Interface

open webui gives you a chatgpt-like interface that runs on your own machine. pair it with NanoGPT and you get 400+ models without giving anyone your data. it's the setup i recommend to everyone who asks me about self-hosted AI.

tl;dr: Open WebUI plus NanoGPT gives you a self-hosted ChatGPT-like interface with 400+ models for $6-8/month. setup takes 5-10 minutes with Docker. your data stays local, conversations are stored on your machine, and you control everything.

Key Takeaways:

Open WebUI + NanoGPT costs $6-8/month vs $20/month for ChatGPT Plus with worse privacy
Docker setup takes 5-10 minutes: one docker run command or docker-compose up
Open WebUI supports RAG document upload, web search, multi-user setup, and runs on a Raspberry Pi 4

i've been running this combo for two months on a mini PC (Intel N100, 16GB RAM). it's my daily driver. total cost: $6-8/month for NanoGPT credits, $0 for the interface.

why open webui

open webui (formerly Ollama WebUI) is the most popular open-source AI chat interface. clean, fast, actively maintained. it's easier to set up than librechat and has a better mobile experience.

feature	open webui	librechat	chatgpt	NanoGPT web
self-hosted	yes	yes	no	no
open-source	yes	yes	no	no
multi-model	yes	yes	limited	yes
mobile UI	excellent	good	good	basic
RAG support	yes	yes	yes	no
image gen	yes	yes	yes	no
user management	yes	yes	n/a	no
setup difficulty	easy	medium	none	none

open webui is easier to set up than librechat and has a more polished mobile experience. librechat has more plugins and customization. pick based on your priorities - or just start with open webui and switch later if you need more features.

what you need

NanoGPT account with API key (sign up here)
Docker Desktop installed (or Docker Engine on Linux)
5-10 minutes

installing Docker

short answer: Mac/Windows: download Docker Desktop. Linux: curl https://get.docker.com | sh then log out and back in.

Mac/Windows: download Docker Desktop

Linux (Ubuntu/Debian):

curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker $USER
# log out and back in

installation

method 1: Docker run (simplest)

short answer: docker run -d -p 3000:8080 with volume mounting. single command, takes 2-3 minutes.

docker run -d -p 3000:8080 \
  --add-host=host.docker.internal:host-gateway \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

method 2: Docker compose (recommended)

short answer: docker-compose.yml with port mapping and volume. docker compose up -d for easy management.

create docker-compose.yml:

version: '3.8'
services:
  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    ports:
   - "3000:8080"
    volumes:
   - open-webui:/app/backend/data
    environment:
   - WEBUI_AUTH=true
    restart: always

volumes:
  open-webui:

then:

docker compose up -d

access open webui

short answer: open http://localhost:3000 in your browser. create your admin account on first visit.

open your browser, go to:

http://localhost:3000

create your admin account on first visit. that's it for installation.

connecting NanoGPT

step 1: open settings

short answer: click profile icon bottom left, go to Settings, then Connections.

click your profile icon (bottom left) → Settings → Connections

step 2: add OpenAI API connection

short answer: under OpenAI API, click plus and fill in the NanoGPT base URL and your API key.

under OpenAI API, click the + button and fill in:

API Base URL: https://api.nano-gpt.com/v1
API Key: your NanoGPT API key

click Save.

step 3: verify connection

short answer: go back to chat. if NanoGPT models appear in the dropdown, you're connected.

go back to the chat interface. in the model dropdown, you should see NanoGPT models listed. if you see models like gpt-4o, claude-3-5-sonnet, etc., you're connected.

step 4: test it

short answer: pick a model and send a test message. if you get a response, setup is complete in about 3 minutes.

pick a model and send:

Hello! Can you confirm which model you are?

if you get a response, setup is complete. took me about 3 minutes total.

model configuration

adding custom models

short answer: go to Settings, Models, Add Model. enter model ID like gpt-4o, base URL, and API key.

if models don't appear automatically:

go to Settings → Models
click Add Model
enter model ID (e.g., gpt-4o), base URL, and API key
save

my recommended model setup

short answer: gpt-4o for general use, Claude 3.5 for writing, GPT-4o-mini for quick questions, DeepSeek for budget, Gemini for long docs.

model	use case	settings
gpt-4o	general, code	default
claude 3.5 sonnet	writing, analysis	default
gpt-4o-mini	quick questions	default
deepseek v3	budget coding	default
gemini 1.5 pro	long documents	default

model presets

short answer: create presets for different tasks: coding at temp 0.3, writing at temp 0.8, quick answers at temp 0.5.

open webui supports presets. create them for different tasks:

start a new chat
select your model
click the gear icon next to the model
set temperature (0.7 balanced, 0.3 focused), max tokens, system prompt
save as preset

i have presets for "coding" (gpt-4o, temp 0.3), "writing" (claude sonnet, temp 0.8), and "quick answers" (gpt-4o-mini, temp 0.5).

advanced features

RAG (document upload)

short answer: upload PDFs and documents, then ask questions about them. Gemini 1.5 Pro is best for this with its huge context window.

open webui supports document upload and retrieval-augmented generation:

click the + button in chat
upload a PDF, text file, or document
ask questions about the document
the model uses the document as context

works with any NanoGPT model that supports long contexts. gemini 1.5 pro is best for this - huge context window, good at document analysis.

web search

short answer: enable web search in settings to let the AI search the web and include results in context with citations.

open webui can search the web and include results in the AI's context:

go to Settings → Web Search
enable web search
configure search provider (SearXNG, Google, etc.)
the AI can now search and cite sources

multi-user setup

short answer: enable user registration in admin settings. each user gets their own history. all share the NanoGPT API key.

for sharing with others:

go to Settings → Admin → Users
enable user registration (or invite manually)
each user gets their own conversation history
all users share the NanoGPT API key and balance

troubleshooting

problem	cause	fix
no models showing	API connection not configured	check Settings → Connections
"invalid API key"	wrong key or URL	verify key at NanoGPT dashboard
slow responses	model or network	try a faster model (gpt-4o-mini)
Docker won't start	port conflict	change port: `-p 3001:8080`
can't access from phone	firewall	open port 3000 in firewall
blank page	browser cache	hard refresh (Ctrl+Shift+R)

Docker commands

short answer: docker ps to check status, docker logs open-webui to debug, docker restart open-webui to restart.

# check status
docker ps

# view logs
docker logs open-webui

# restart
docker restart open-webui

# update to latest version
docker pull ghcr.io/open-webui/open-webui:main
docker stop open-webui
docker rm open-webui
# re-run the original docker run command

# or with compose
docker compose pull
docker compose up -d

testing your API key

short answer: curl the NanoGPT models endpoint with your key. if it returns model data, the key works and the issue is in Open WebUI config.

curl https://api.nano-gpt.com/v1/models \
  -H "Authorization: Bearer YOUR_KEY_HERE"

if this returns model data, the issue is in open webui configuration. if it errors, the issue is with your NanoGPT account or key.

open webui vs librechat

i've used both extensively. here's my honest comparison:

aspect	open webui	librechat	winner
setup ease	easier	harder	open webui
UI polish	better	good	open webui
mobile experience	excellent	decent	open webui
plugin ecosystem	growing	larger	librechat
customization	good	excellent	librechat
multi-user	good	better	librechat
documentation	good	better	librechat
update frequency	weekly	weekly	tie

my pick: open webui for personal use. librechat for teams or heavy customization needs. see our librechat setup guide for the alternative.

accessing from your phone

if open webui is running on your local network, access it via your computer's IP address:

http://192.168.x.x:3000

for remote access, set up a VPN or reverse proxy with HTTPS. don't expose port 3000 directly to the internet.

the mobile UI is genuinely good. i use it on my phone regularly.

running on a raspberry pi

yes, it works on a raspberry pi 4 or 5 with 4GB+ RAM. the UI is lightweight - the heavy lifting happens on NanoGPT's servers. i ran it on a pi 4 for a while before switching to a mini PC.

data privacy

open webui runs locally. your conversations are stored on your machine. the only external connection is to NanoGPT's API when you send a message. open webui doesn't phone home or collect analytics.

for maximum privacy, run it on a local machine (not a VPS). if you need remote access, use a VPN.

updating

docker compose pull
docker compose up -d

check for updates weekly or so. the project updates frequently.

my experience

two months running open webui + NanoGPT on a mini PC. daily driver for AI chat.

pros: fast UI, great mobile experience, easy model switching, my data stays local.

cons: occasional Docker issues after updates, RAG setup requires extra configuration.

total cost: $6-8/month for NanoGPT credits. compared to $20/month for chatgpt plus with worse privacy, it's a clear win.

👉 Start with NanoGPT

Last updated: July 2026

NanoGPT + LibreChat Setup - alternative interface
NanoGPT API Key Setup - get your key first
Best NanoGPT Models - full model list
NanoGPT Pricing - cost breakdown
NanoGPT Privacy Review - data handling
SillyTavern + NanoGPT - for roleplay users

Disclosure: This article contains affiliate links. If you sign up through our referral link, you get a 5% discount and we earn a small commission. This doesn't affect our reviews - we pay for all services ourselves.