Install

Bytebot runs entirely in Docker. There is no npm package or pip install. The canonical path is Docker Compose (Option 1). Railway is a one-click cloud alternative (Option 2).

Option 1: Docker Compose (self-hosted, recommended)

Prerequisites: Docker and Docker Compose installed, plus an AI provider API key.

Step 1 -- Clone the repo

git clone https://github.com/bytebot-ai/bytebot.git
cd bytebot

Step 2 -- Set your AI provider key

Choose one provider and write the key to docker/.env:

# Anthropic Claude (recommended)
echo "ANTHROPIC_API_KEY=sk-ant-..." > docker/.env

# OpenAI GPT
# echo "OPENAI_API_KEY=sk-..." > docker/.env

# Google Gemini
# echo "GEMINI_API_KEY=..." > docker/.env

Step 3 -- Start all services

docker-compose -f docker/docker-compose.yml up -d

Services are ready in roughly 2 to 3 minutes on first run (image pull time varies). On subsequent starts it is faster.

Access points after startup

Service	URL
Tasks UI	http://localhost:9992
Agent API	http://localhost:9991
Desktop control API	http://localhost:9990

Useful management commands

# Tail agent logs
docker-compose -f docker/docker-compose.yml logs -f bytebot-agent

# Stop all services
docker-compose -f docker/docker-compose.yml down

# Reset everything (removes volumes)
docker-compose -f docker/docker-compose.yml down -v

Option 2: Railway (one-click cloud deploy)

Click the Railway deploy button on the Bytebot GitHub README or bytebot.ai, then add your AI provider API key in the Railway environment variables panel. No local Docker required.

Railway costs apply based on resource usage. This is not a Bytebot fee -- it is a Railway infrastructure fee.

No Template or Boilerplate

Bytebot does not ship a separate starter template or boilerplate repo. The GitHub repository itself (bytebot-ai/bytebot) is the canonical starting point; clone it directly as shown above.