Creating a Prompt

A Prompt defines the system instructions given to an agent. Prompts are managed separately from agents so you can reuse the same prompt across multiple agents or swap prompts without changing agent configuration.

Creating a Prompt

1. Go to App Panel → your app → AI → Prompts.

2. Click Create Prompt.

   

3. Fill in the prompt details.

   

   Field	Description
   Name	A unique identifier for this prompt (e.g. patient-summary-system)
   Type	system — injected before the conversation; user — sent as a user turn
   Description	Internal note explaining what this prompt is for
   Prompt Content	The instruction text sent to the model
   Version	Auto-incremented each time the prompt is saved; new versions start as drafts

   Both system and user type prompts support variables using double-brace syntax (e.g. {{patient_name}}). Variables are extracted automatically and passed at runtime — system_variables= for system prompts, variables= for user prompts.

   

4. Click Create Prompt.

   

5. The prompt appears in the list and is ready to be attached to an agent.

Editing a Prompt

Click the prompt name in the list to open the edit view. Each save creates a new draft version — changes only take effect for agents once you activate the new version.

A side-by-side diff is shown for each edit so you can review exactly what changed.

Prompt Versioning

Every save creates a new version, which starts as a draft. Drafts are not used by agents until you explicitly promote one by clicking Make Active. Only the active version is resolved when an agent references this prompt by name.

note

Activating a new version immediately affects every agent attached to this prompt. To experiment safely, create a separate prompt with a different name and test it on a non-production agent first.

Next Steps

With prompts ready, define tools your agent can call.