Multiplayer Tests
Our automation framework allows you to write and execute multiplayer tests, enabling multiple AI agents to play your game simultaneously. This is essential for testing functionality that requires player interaction, such as lobbies, friend systems, trading, cooperative gameplay, and competitive features.
Writing a Multiplayer Test
To write a multiplayer test, we first need to add more players to a test. This can be done by clicking the “Add Player” button in the test editor action bar.
Each player represents a separate game instance controlled by an independent AI agent. You can add as many players as your project has agents available, though keep in mind that:
- More players increase test complexity and maintenance overhead
- Additional players require more computational resources (fewer other tests can run simultaneously + you need more agent hours for the multiplayer test)
- Coordination between many players can be challenging to manage
Step Synchronization
All players in a multiplayer test must have the same number of steps. The AI agents execute steps synchronously, meaning all players complete Step 1 before any player moves to Step 2, and so on.
This synchronization ensures that:
- Players can interact with each other reliably
- Test outcomes are predictable and reproducible
In this example, we have a multiplayer test with two players. To ensure that
the second player only tries to claim the item after the other player
started sharing it, we synchronize the steps. In step 1, the first player
initiates an item trade. In step 2, the second player interacts with player
1 to receive the item. This way, we can ensure that the second player only
tries to receive the item after player 1 has initiated the trade. 
Player Communication
Players can communicate through a shared message channel using broadcast functions. This allows agents to:
- Exchange information (lobby codes, friend IDs, coordinates)
- Coordinate timing (“ready to start”, “waiting at location”)
- Share game state updates (“collected the item”, “reached checkpoint”)
To demonstrate player communication, we can create a simple test where two players exchange messages. The first player sends a message with a list of pokemons, and the second player chooses their favorite pokemon from that list and sends back a response.
Here we can see the agents messages in the live chat of the test run. The first player sends a list of pokemons, and the second player responds with marshtomp as their favorite pokemon.
