> ## 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.
# Personas
> Who calls your agent, and how they sound and act.
A persona is a reusable caller: their personality, voice and call conditions. Give a [Digital Human](https://docs.getbluejay.ai/test/simulations/digital-humans.md#runs-as) several personas to check that your agent handles the same test with every kind of caller.
## How personas run
* A Digital Human runs once as each persona it lists, and once as the baseline, its own settings, when it has none.
* A persona replaces only the caller fields it sets. The intent, success criteria and everything else come from the Digital Human, so every call still tests the same goal.
* Editing a persona changes every Digital Human that runs as it, from the next run on. Past conversations keep how they ran.
* Each conversation records the persona it ran as.
## Fields
| Field | Write or set |
| - | - |
| Name | Who this caller is, for example "Rushed commuter". Names are unique in a workspace. |
| Personality Description | Who the caller is and their attitude, in one short line, for example "retiree on a family plan, hard of hearing". |
| Voice | Language, accent and gender, or a custom voice. |
There is no emotion setting. To make the caller angry, rushed or confused, write that attitude into **Personality Description**.
## Behavior
Use behavior settings to test whether your agent copes with harder callers, for example a fast talker on a bad connection.
* **Patience (seconds)** is how long the caller waits before it replies.
* **No Response Timeout (seconds)** must be 15 seconds or more.
* **Creativity** from 0.0 to 0.3 makes the caller say much the same thing on each run.
* **Speaks First**: the caller speaks first unless you change it. Some voice platform connections start with the agent.
## Scripted responses
A scripted response makes a reply the same on each run. Each one pairs a line the agent says with the caller's reply: words, keypad tones, silence, or an audio clip. Scripted responses live on the Digital Human, so they apply whichever persona it runs as. The Persona API accepts `scripted_responses` too, but runs use the Digital Human's.
* Under **When the agent says**, write a line the agent says, for example "Can I have your account number?". A description of agent behavior never matches.
* Matching on context also catches lines with the same meaning in other words.
* If **How scripted responses fire** runs them in order, each one fires once, in list order.
## The Default persona
Each workspace has a **Default persona**, a plain caller with an American voice and no background noise. A plan uses it when you do not pick a persona. You can edit the Default persona, but not delete it.
## Add personas
Open **Digital Humans**, go to the **Personas** tab, and click **New persona**, or upload a [personas CSV](https://docs.getbluejay.ai/test/simulation-types/upload-digital-humans.md#personas-columns). A [simulation plan](https://docs.getbluejay.ai/test/simulations/create.md) also picks personas for you and saves them here when you run.
Deleting a persona removes it from every Digital Human that runs as it.
## Write a good persona
* Describe the person, never the task. "Cancel my plan" belongs in the Digital Human.
* Make each persona differ from the others in one way that matters to your agent, such as a strong accent, a noisy line, or an impatient caller.
* Keep personas few and reuse them. Each persona on a Digital Human is one more call per run.
## Related
* [Digital Human Reference](https://docs.getbluejay.ai/test/simulations/digital-humans.md)
* [Personas API](https://docs.getbluejay.ai/api-reference/endpoint/list-personas.md)
* [Persona MCP tools](https://docs.getbluejay.ai/mcp/testing-tools.md#personas)
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.