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

# Prompts library

It's a central place to create, manage, and reuse prompt templates across your agents and workflows — with version tracking, response testing, and reference lookup built in.

Navigation: select a project, then go to **Resources** > **Prompt Library**.

## Prompt lifecycle

A typical prompt lifecycle in Prompt Library:

```mermaid actions={false} theme={null}
%%{init: {'theme': 'base', 'themeVariables': {'background': 'transparent', 'primaryColor': '#EAF8E4', 'primaryTextColor': '#1a1a1a', 'primaryBorderColor': '#5CC83A', 'lineColor': '#5CC83A', 'secondaryColor': 'transparent', 'tertiaryColor': 'transparent', 'clusterBkg': 'transparent', 'clusterBorder': '#5CC83A', 'titleColor': '#1a1a1a', 'clusterLabelBackground': 'transparent', 'edgeLabelBackground': '#EAF8E4'}}}%%
flowchart LR
    A[Create Prompt] --> B[Test and Compare]
    B --> C{Satisfactory?}
    C -- No --> D[Refine and Version]
    D --> B
    C -- Yes --> E[Activate Prompt]
    E --> F[Reuse and Manage]
```

| Stage | Description |
| - | - |
| Create Prompt | Create a reusable prompt template and define its instructions and content. |
| Test and Compare | Test the prompt across one or more models and compare responses, token usage, and response times. |
| Refine and Version | Update the prompt based on test results and save new versions as needed. |
| Activate Prompt | Promote a validated version to active status for production use. |
| Reuse and Manage | Reuse the prompt across the project and manage it through version history and references. |

## Create and manage prompts

The Prompt Library page lists all prompts in the selected project. For each prompt, you can:

* Define prompt instructions and content.
* Add tags for categorization.
* Save prompts as drafts.
* Promote validated prompts to the active version.

![Create prompts](https://mintcdn.com/koreai/NwgdJQN5p8NwI2Qj/agent-platform/images/prompts-create.png?fit=max\&auto=format\&n=NwgdJQN5p8NwI2Qj\&q=85\&s=749e301891b519e940544389b6e28d22)

## Manage prompt versions

Each prompt supports multiple versions. Use the **Versions** tab to:

* View all available versions.
* Track revisions over time.
* Compare changes between versions.
* Promote a selected version to active status.

### Prompt states

| State | Description |
| - | - |
| Draft | A work-in-progress version used for editing and testing. |
| Active | The published version available for production use. |

## Test and compare prompts

Use **Test & Compare** to evaluate prompt behavior across multiple language models before deployment. You can:

* Select a prompt version.
* Choose models for comparison.
* Provide a test user message.
* Upload an optional context file.
* Run the prompt against selected models simultaneously.

![Test prompts](https://mintcdn.com/koreai/IKjI4mL70tZk-BY8/agent-platform/images/prompts-test.png?fit=max\&auto=format\&n=IKjI4mL70tZk-BY8\&q=85\&s=249b71338c647be9b59f3d468c3bffc8)

Results appear side by side with response metrics including response time and token usage.

### Use context during testing

You can upload files to include additional context during a test run. The uploaded content is available during prompt execution, letting you evaluate how the prompt handles real-world, project-specific information.

## View prompt references

The **References** tab shows where a prompt is used within the project. Use it to:

* Understand prompt dependencies.
* Assess the impact of changes before modifying an active prompt.
* Identify agents and workflows that rely on the prompt.

## Use a prompt in agents

A Prompt Library prompt linked to an agent replaces the agent's generated system prompt. Keep the prompt complete, including the agent's limitations and rules. Don't store a fragment.

* Pin one prompt version to each agent that uses it.
* A prompt change doesn't update a deployed agent. Publish a new agent version, redeploy, and test in a new session.

## Best practices

* Use descriptive names that clearly indicate a prompt's purpose.
* Add tags to improve organization and searchability.
* Save changes as drafts while iterating.
* Test prompts across multiple models before promoting to active status.
* Review references before modifying active prompts.
* Use version history to track prompt evolution and preserve previous revisions.

***


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