Cookbook

Build Reliable AI Agents With Simulated Environments

A collection of reference agents demonstrating how to build, test, and deploy AI agents using the Veris simulation platform — or run them standalone.

Each example is a complete, working agent with its own README, test suite, and Veris simulation scenarios. Use them as starting points for your own agents or as references for integrating with Veris.

Each agent can run independently — just follow the README in its directory. You'll typically need:

1. Clone the repo and cd into the agent directory
2. Install dependencies (uv sync or pip install)
3. Set environment variables (API keys, etc.)
4. Start the server (uvicorn or docker compose up)

Veris provides sandboxed environments with simulated users and services so you can test your agent end-to-end before deploying to production.

1. Install the Veris CLI

uv tool install veris-cli
veris login

2. Create an environment

cd <agent-directory>
veris env create

3. Configure and push

Update .veris/veris.yaml with your environment variables, then:

veris env push

4. Run simulations

Generate test scenarios and run them:

veris scenarios create --num 25
veris run

Each agent includes pre-built scenarios in its scenarios/ directory that you can also run directly.

For the full walkthrough, see the Veris documentation.

cookbook/
  bca-agent/               Banker Connections Agent (Google ADK)
  card-replacement-agent/   Card Replacement Agent (OpenAI Agents SDK)
  pm-analyst/               PM Analyst (Google ADK)
  procurement-agent/        Procurement Agent (OpenAI Agents SDK)

Each agent follows a consistent layout:

<agent>/
  app/              Application code and agent logic
  tests/            Unit and integration tests
  scenarios/        Veris simulation scenarios
  .veris/           Veris sandbox configuration
  README.md         Setup and usage instructions

To add a new cookbook example:

1. Create a new directory at the repo root
2. Include a README.md with setup instructions, architecture overview, and Veris deployment steps
3. Add a .veris/ directory with sandbox configuration
4. Include at least one scenario in scenarios/
5. Open a pull request

• Veris Documentation — Full platform docs, quickstart, and API reference
• Veris CLI — CLI installation and commands
• OpenAI Agents SDK — Agent framework used by card-replacement and procurement agents
• Google ADK — Agent framework used by BCA and PM Analyst