Skip to main content
Get OneRun running locally and create your first agent evaluation in minutes.

​
Prerequisites

  • Docker installed on your machine
  • Anthropic API key for running AI agents

​
Step 1: Start Temporal Server

OneRun uses Temporal for orchestrating simulations. First, start the Temporal server: 
# Clone the OneRun repository
git clone https://github.com/onerun-ai/onerun.git
cd onerun

# Navigate to temporal directory and start services
cd docker/temporal
docker compose up
Keep this terminal running. Temporal needs to stay active for OneRun to work.

​
Step 2: Configure Environment

In a new terminal, set up your environment configuration: 
# Navigate to docker directory
cd docker

# Copy environment template
cp .env.example .env

# Edit the .env file with your API key
# Replace YOUR_ANTHROPIC_API_KEY with your actual key
Update your .env file: 
ANTHROPIC_API_KEY=your-actual-anthropic-api-key-here
Support for other AI models and better configuration management is coming in future releases.

​
Step 3: Start OneRun Platform

# From the docker directory
docker compose up
Your OneRun instance will be available at http://localhost:3000

​
Step 4: Create Your First Evaluation

  1. Open OneRun in your browser at http://localhost:3000
  2. Sign Up for a new account or log in if you already have one
  3. Create a Project to organize your evaluations
  4. Define an Agent that represents your AI system
  5. Set Objectives to measure what success looks like
  6. Create a Simulation to test specific scenarios
OneRun will automatically generate diverse personas and orchestrate conversations once your worker is running.

​
Step 5: Run Your Worker

To handle actual conversations, you need to run a worker that implements your agent logic: 
# Navigate to the Python SDK examples
cd python-sdk/examples

# Install dependencies (requires uv)
uv pip install .

# Set up environment variables
cp .env.example .env

# Edit .env with your OneRun configuration
# ONERUN_API_BASE_URL=http://localhost:3000
# ONERUN_API_KEY=your-api-key-from-onerun-ui
# ONERUN_PROJECT_ID=your-project-id (found in Project Settings page)
# ONERUN_AGENT_ID=your-agent-id (found in Agent Details page)
# ANTHROPIC_API_KEY=your-anthropic-key

# Start the worker
uv run multi-turn-agent/main.py
Your worker will:
  • Poll for assigned conversations during active simulations
  • Handle turn-by-turn exchanges with generated personas
  • Use your AI model to generate agent responses
  • Report results back to OneRun for evaluation

​
Next Steps

Now that OneRun is running, explore these key concepts:

Core Concepts

Understand projects, agents, and objectives

Workers

Learn how to integrate your AI agent

Use Cases

See how teams use OneRun in production

API Reference

Explore the complete API documentation
Your OneRun instance is completely self-hosted. All your evaluation data stays on your infrastructure with no external dependencies except for AI model API calls.