Install Presenton

Choose the installation that matches where Presenton will run.

Installation	Best for	Data location
Docker Run	The fastest self-hosted setup	A host folder mounted at `/app_data`
Docker Compose	Repeatable servers and environment-based configuration	`./app_data` beside the Compose file
Desktop app	A native app without managing containers	The application data directory
GPU-enabled Docker	Local AI workloads that use an NVIDIA GPU	A host folder mounted at `/app_data`

The default self-hosted URL is http://localhost:5001. Port 5001 on the host maps to port 80 in the container.

Option 1: Docker Run

Install Docker Desktop or Docker Engine, then start the current v0.9.0-beta image.

Linux and macOS
Windows PowerShell

mkdir -p app_data

docker run -it --name presenton \
  --publish 5001:80 \
  --volume "./app_data:/app_data" \
  ghcr.io/presenton/presenton:v0.9.0-beta

New-Item -ItemType Directory -Force app_data

docker run -it --name presenton `
  --publish 5001:80 `
  --volume "${PWD}\app_data:/app_data" `
  ghcr.io/presenton/presenton:v0.9.0-beta

The versioned tag keeps this guide reproducible. Replace it with ghcr.io/presenton/presenton:latest when you intentionally want the newest published image.

Publish only the ports you need

Port mapping	Required when
`5001:80`	Always for the default browser, REST API, and MCP URL
`1455:1455`	Only when signing in with an existing ChatGPT/Codex subscription

To use Sign in with ChatGPT, publish the OAuth callback port and select the Codex provider:

docker run -it --name presenton \
  --publish 5001:80 \
  --publish 1455:1455 \
  --env LLM=codex \
  --env CODEX_MODEL=gpt-5.5 \
  --volume "./app_data:/app_data" \
  ghcr.io/presenton/presenton:v0.9.0-beta

The browser returns authentication to http://localhost:1455/auth/callback. Without 1455:1455, ChatGPT/Codex sign-in cannot complete inside the Docker container. API-key providers do not need this extra port.

Run in the background

Add --detach to the command, then manage the container with:

docker logs --follow presenton
docker stop presenton
docker start presenton

Use another host port

Change only the value before the colon. This example opens Presenton at http://localhost:8080:

docker run -it --name presenton \
  --publish 8080:80 \
  --volume "./app_data:/app_data" \
  ghcr.io/presenton/presenton:v0.9.0-beta

Option 2: Docker Compose

Clone the source repository when you want a repeatable deployment with a .env file and the repository’s production profiles.

git clone https://github.com/presenton/presenton.git
cd presenton

Create a .env file beside docker-compose.yml:

PRESENTON_HTTP_HOST_PORT=5001
AUTH_USERNAME=admin
AUTH_PASSWORD=replace-with-a-long-password

LLM=openai
OPENAI_API_KEY=replace-with-your-key
OPENAI_MODEL=gpt-4.1

IMAGE_PROVIDER=pexels
PEXELS_API_KEY=replace-with-your-key
CAN_CHANGE_KEYS=false

Start the production service:

docker compose up production

Run it in the background with docker compose up --detach production. To use port 8080 without editing .env:

PRESENTON_HTTP_HOST_PORT=8080 docker compose up production

The repository provides four services:

Service	Purpose
`production`	Normal self-hosted deployment
`production-gpu`	Production deployment with NVIDIA GPU access
`development`	Source-mounted development environment
`development-gpu`	Development environment with NVIDIA GPU access

Compose already publishes 1455:1455 for the Codex OAuth callback. This is what lets users select Sign in with ChatGPT and use an eligible existing ChatGPT/Codex subscription without an OpenAI API key.

Option 3: Desktop app

Use the desktop application when you want a native installation and do not need to expose the REST API or MCP server to other tools.

Platform	Package
macOS Apple Silicon	`.dmg` for arm64
Windows	`.exe` for x64
Linux	`.deb` for x64

Install the desktop app

Download the correct package, complete the operating-system prompts, and connect a provider.

The Electron desktop app does not expose the MCP server. Use Docker or another web deployment when you need MCP or a remotely accessible API.

Option 4: NVIDIA GPU

Install the NVIDIA driver and NVIDIA Container Toolkit, confirm docker run --gpus all works, then use either the GPU Compose service or --gpus=all.

docker run -it --name presenton \
  --gpus=all \
  --publish 5001:80 \
  --env LLM=ollama \
  --env OLLAMA_MODEL=llama3.2:3b \
  --volume "./app_data:/app_data" \
  ghcr.io/presenton/presenton:v0.9.0-beta

Or with Compose:

docker compose up production-gpu

GPU access helps compatible local workloads; it does not make hosted API providers run locally.

Complete the first-run setup

Open Presenton

Visit http://localhost:5001, or the host port you selected. For a remote server, use its HTTPS address.

Choose the text model

Select the provider that creates outlines, slide content, and AI edits. You can configure it in Settings or with environment variables.

Choose images and search

Connect a stock or generated-image provider. Enable web search only when the presentation needs current public information.

Verify persistence

Create a small presentation, restart the container, and confirm the presentation remains available. If it disappears, inspect the /app_data mount.

Secure shared deployments

Configure the single admin account, use HTTPS through a reverse proxy, and keep secrets in a protected .env file or secret manager.

Verify the installation

docker ps --filter name=presenton
docker logs --tail 100 presenton

Open the UI and generate a short two-slide test deck. A successful test confirms that the application, selected text provider, image configuration, and persistent storage are working together.

Keep the /app_data mount when replacing or upgrading a container. Starting a new container without it creates an empty instance and makes the existing presentations and settings appear missing.

Configure the instance

Add providers, authentication, search, memory, and storage settings.

Create your first presentation

Generate, edit, and export a complete deck.

Configure by goal

See exactly which ports and variables enable ChatGPT/Codex sign-in, local models, images, web search, API access, MCP, authentication, and external storage.

Get started

Learn the basics

Features

Tutorials

API guides

Deploy and configure

Develop and contribute

This release

Option 1: Docker Run

Publish only the ports you need

Run in the background

Use another host port

Option 2: Docker Compose

Option 3: Desktop app

Install the desktop app

Option 4: NVIDIA GPU

Complete the first-run setup

Verify the installation

Configure the instance

Create your first presentation

Configure by goal

​Option 1: Docker Run

​Publish only the ports you need

​Run in the background

​Use another host port

​Option 2: Docker Compose

​Option 3: Desktop app

Install the desktop app

​Option 4: NVIDIA GPU

​Complete the first-run setup

​Verify the installation

Configure the instance

Create your first presentation

Configure by goal

Option 1: Docker Run

Publish only the ports you need

Run in the background

Use another host port

Option 2: Docker Compose

Option 3: Desktop app

Option 4: NVIDIA GPU

Complete the first-run setup

Verify the installation