> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getbluejay.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Create a Simulation

> Create a simulation.

A simulation calls your agent with Digital Humans, which are realistic test customers, and scores each conversation.

<video controls muted playsInline preload="metadata" src="https://mintcdn.com/bluejay/eKcFBkjZX47pI8Ss/videos/create-simulation-dialog.mp4?fit=max&auto=format&n=eKcFBkjZX47pI8Ss&q=85&s=5f884e3051bdd9bba7f40bc0c0ff13eb" />

## Connect your agent

**1. Add the agent**

Switch to the [workspace](/administration/workspaces.md#what-a-workspace-holds) you will test in, click **+** in the sidebar header, and connect your agent the way your customers reach it. [Connect Your Agent](/integrations/overview.md) lists the setup page for each voice, chat, and email connection.

**2. Say who starts the conversation**

Choose **Inbound** if your customers contact the agent, or **Outbound** if the agent contacts them. For an outbound voice or text agent, add an [outbound trigger webhook](/test/simulations/reference.md#outbound-trigger-webhook) so Bluejay can tell your agent whom to call or message.

Paste the agent's system prompt too. Bluejay uses it to write Digital Humans and metrics that fit your agent.

Testing an email agent? Make sure its inbox [accepts mail from Bluejay](/integrations/email.md#allow-bluejays-mail) first.

## Create a simulation

**1. Choose a type**

Open **Simulations** in the sidebar, click **Create**, and click a **Simulation Type** card.

| Card | Use it to |
| - | - |
| [**Generate from Natural Language**](/test/simulation-types/natural-language.md) | Make Digital Humans from a description. |
| [**Goal Adherence**](/test/simulation-types/goal-adherence.md) | Make Digital Humans from your agent's goals. |
| [**Generate from Knowledge Base**](/test/simulation-types/knowledge-base.md) | Make Digital Humans from a Knowledge Base. |
| [**Generate from System Prompt**](/test/simulation-types/system-prompt.md) | Make Digital Humans from your agent's prompt. |
| [**Replay From Transcript**](/test/simulation-types/replay-from-transcript.md) | Replay real conversations against your agent. |
| [**Generate From Workflow**](/test/simulation-types/generate-from-workflow.md) | Test each path through your agent's multi-agent workflow. |
| [**Customer Journeys**](/test/simulation-types/customer-journeys.md) | Test how your agent treats returning customers. |
| [**Upload Digital Humans**](/test/simulation-types/upload-digital-humans.md) | Make Digital Humans from a spreadsheet. |
| [**Use Previous Digital Humans**](/test/simulation-types/previous-digital-humans.md) | Rerun the same Digital Humans across tests. |
| [**Load Testing**](/test/load-testing/overview.md) | Test your agent under many conversations at once. |
| [**Voicemail**](/test/simulation-types/voicemail.md) | Test how your agent handles voicemail. |
| [**IVR Flow**](/test/simulation-types/ivr-flow.md) | Test how your agent gets through a phone menu. |
| [**Scenario Adherence**](/test/simulation-types/scenario-adherence.md) | Test that your agent follows a required flow. |

**2. Generate and run**

Click **Save** and **Generate Digital Humans**, then review the list, because each Digital Human places one conversation. Click **Run** to start the calls.

To test the same callers again after you change the agent, open the simulation and click **Create New Run**.

## Read the results

Start with the failed conversations to see which metric failed and where it went wrong. Bluejay groups similar failures into [issues](/test/issues.md), so you can fix the problem behind many failed conversations at once. To see what your agent did behind the transcript, such as each model call, send [traces](/test/traces.md).

![A run's results page with the success rate chart and conversation rows, and a failed conversation's details panel open showing its success criteria and failure reason](https://mintcdn.com/bluejay/eKcFBkjZX47pI8Ss/images/simulation-results.png?fit=max\&auto=format\&n=eKcFBkjZX47pI8Ss\&q=85\&s=e673132991a926166306e31971df50c1)

If a conversation did not finish as **Completed**, look up its code in [Error codes](/test/simulations/reference.md#error-codes). After you fix the problem, click **Rerun** and choose **Rerun failed**.

To follow your metrics from run to run, chart them on a [dashboard](/test/dashboards.md).

## Let Bluejay AI do it

You can also ask [Bluejay AI](/improve/bluejay-ai.md) to create and run the simulation for you. For example:

* "Run a simulation on my agent."
* "Create a simulation with 10 frustrated customers who want to cancel their subscription."

Bluejay AI shows you the Digital Humans it made and asks before it starts the run. If you have test cases in a spreadsheet, attach the `.csv` or `.xlsx` file and ask it to turn them into a simulation.

## Related

* [Simulation Reference](/test/simulations/reference.md)
* [Digital Human Reference](/test/simulations/digital-humans.md)
* [Metrics](/test/metrics.md)
* [Bluejay AI](/improve/bluejay-ai.md)
* [Simulation API](/api-reference/endpoint/create-simulation.md)
* [Simulation MCP tools](/mcp/testing-tools.md#simulations)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.