AgentPlayground ships as a downloadable package that runs on your own computer through Docker. No
cloud account required to install it — you bring your own AI provider key (or run fully local
models with Ollama) once it's up.
- Docker Desktop 4.x or later — download at docker.com/products/docker-desktop
- 4GB RAM minimum (8GB recommended if you plan to run local models with Ollama)
- 5GB free disk space
- Install Docker Desktop, if you don't already have it, and make sure it's running (look for the whale icon in your taskbar).
- Download the latest release from agentplayground.net/download and extract the ZIP to any folder on your computer.
- Start the app:
- Windows: double-click
start.bat - Mac/Linux: open a terminal in that folder and run
./start.sh
- Your browser opens to
http://localhost:3000. The setup wizard walks you through the rest — see First-run setup for what each step does.
Don't have an API key yet? Get a completely free one at build.nvidia.com
(NVIDIA — free cloud models, no credit card), or a paid one at
platform.openai.com/api-keys (OpenAI) or
console.anthropic.com (Anthropic).
Run stop.bat (Windows) or ./stop.sh (Mac/Linux), or just stop it from Docker Desktop.
To start the app with a bundled local model runtime:
docker compose -f docker-compose.yml -f docker-compose.ollama.yml up -dThen select Ollama in the setup wizard. This needs 8GB+ RAM and can be slow on older machines,
but it's fully free and your data never leaves your computer.
Go to Settings → API Keys inside the app at any time — you don't need to redo setup.
"Docker not found" — make sure Docker Desktop is open and running.
First start is slow — Docker needs to download images on first run (about 500MB). This can take
a few minutes on a slower connection.
App doesn't open automatically — wait 30–60 seconds and visit
http://localhost:3000 manually.
Port 3000 already in use — edit docker-compose.yml, change "3000:3000" to "3001:3000",
then access the app at port 3001.
For more detail on specific errors, see API key errors and
Docker / build issues.