Skip to main content

​
What is the Library?

Your Library is your personal collection of everything you have installed from the Marketplace. It holds:
  • AI Agents you have added or purchased
  • Project Bases you have installed for creating new projects
  • Forked Agents you have customized from open-source originals
  • Custom Agents you have built from scratch
When you install an item from the Marketplace, a UserPurchasedAgent or UserPurchasedBase record is created in the database, linking your user account to that item with details about the purchase type (free, purchased, or subscription) and activation status.

​
Accessing Your Library

1

Open Library

Click Library in the sidebar navigation.
2

Browse Categories

Switch between the Agents and Bases tabs to view your installed items.
3

Manage Items

Enable, disable, edit, fork, or remove items from your collection.

​
Managing Agents

​
Enabling and Disabling

Control which agents appear in your project’s chat agent selector:
  • Enabled agents: Show up in the agent dropdown within every project
  • Disabled agents: Hidden from the selector but not removed from your library
  • Toggle: Click the switch on any agent card to enable or disable it
Disabling an agent does not delete it or cancel a subscription. It only hides the agent from the chat selector. You can re-enable it at any time.

​
Agent Selection in Projects

When you open a project and use the chat, you pick an agent from a dropdown. The agent selection is persisted to localStorage per project slug, so each project remembers which agent you were using last.

​
Editing Open-Source Agents

For agents with source_type = "open", you can customize the following:

System Prompt

Edit the instructions that define the agent’s behavior, coding style, and constraints

AI Model

Change which AI model the agent uses (for example, switch from GPT-4o to Claude Sonnet)

Name and Icon

Customize the display name and icon shown in the agent selector

Tool Configuration

Adjust which tools the agent can access and how they are described in the system prompt

​
Forking Agents

Forking creates a personal copy of an open-source agent that you can modify without affecting the original.
1

Find a Forkable Agent

In your Library or the Marketplace, look for agents marked as open source with the fork option available.
2

Click Fork

A new agent is created in your library with the same system prompt, tools, and configuration as the original.
3

Customize

Edit the system prompt, change the model, rename it, or adjust tool configurations to match your workflow.
4

Save

Your forked agent is now available in the agent selector for all your projects.
The forked agent stores a parent_agent_id referencing the original, and your user_id as the forked_by_user_id. Forked agents are private by default (is_published = false).

​
Model Override

For open-source agents, you can set a selected_model override in your UserPurchasedAgent record. This lets you use a different AI model than the one the agent creator specified, without forking the entire agent.

​
Managing Bases

​
Installed Bases

View all bases you have added to your library:
  • Preview: See the base’s tech stack, category, and feature list
  • Download count: How many users have installed this base
  • Rating: Average community rating
  • Remove: Delete from your library (does not affect projects already created from this base)

​
Creating Projects from Bases

1

Go to the Dashboard

Navigate to your main Dashboard page.
2

Click Create New Project

Open the project creation dialog.
3

Select Base as Source

Choose Base and pick from your installed bases.
4

Name and Create

Enter a project name and click Create. The base repository is cloned into your new project.

​
Creating Custom Agents

​
Building from Scratch

1

Open Create Agent

In your Library, click Create New Agent.
2

Choose Agent Type

Select the execution strategy:
  • StreamAgent: Real-time streaming with tool calls (best for UI work and fast iteration)
  • IterativeAgent: Multi-step planning and execution (best for complex logic and multi-file changes)
3

Configure the Agent

Set the name, icon, description, AI model, and system prompt.
4

Define Tools

Select which tools the agent can access: file reading/writing, shell commands, web fetching, and more.
5

Test

Try the agent in a test project. Iterate on the system prompt until the output matches your expectations.

​
Writing Effective System Prompts

The system prompt is the most important part of a custom agent. It determines how the agent responds, what coding patterns it follows, and what constraints it respects.

Be Specific

Define exact behavior: “Always use TypeScript. Always add error handling to async functions. Use Tailwind for styling.”

Include Examples

Show the agent what good output looks like. Paste a sample component or function as a reference.

Set Constraints

Specify what the agent should avoid: “Never use inline styles. Do not install new dependencies without asking.”

Define Context

Explain the agent’s role: “You are a senior React developer building a SaaS dashboard. The project uses React 19, TypeScript, and Tailwind CSS.”

​
Publishing to the Marketplace

Share your custom or forked agents with the community:
1

Polish the Agent

Make sure the system prompt produces consistent, high-quality output. Write a clear description and add relevant tags.
2

Click Publish

Open the agent settings in your Library and click Publish to Marketplace.
3

Set Details

Configure category, tags, and pricing model (free, monthly, API, or one-time). Set a price if applicable.
4

Submit

Your agent becomes available in the marketplace. If you chose a paid pricing model, configure your Stripe Connect account to receive payouts.

​
Pricing Your Agent

No charge. Good for building reputation and getting feedback. Most community agents start here.
Published agents are visible to all marketplace users. Only publish agents you are comfortable sharing publicly. You can unpublish at any time by setting is_published = false.

​
Managing AI Models

​
Model Configuration

Control which AI models your agents use:
Set default models for each agent type:
  • StreamAgent default model
  • IterativeAgent default model
  • Per-agent model overrides

​
Disabled Models

You can hide specific models from the chat selector using the disabled_models preference in your user settings. This keeps the model dropdown clean and focused on the models you actually use.

​
Library Organization

Filter your library by:
  • Source type (open-source or closed)
  • Enabled or disabled status
  • Agent type (StreamAgent, IterativeAgent)
  • Search by name or description
Sort by:
  • Recently used
  • Alphabetical
  • Date added
  • Rating
Star frequently used agents for quick access. Starred items appear at the top of the agent selector dropdown.

​
Best Practices

Disable agents you are not actively using so the chat selector stays manageable. Remove bases you no longer need.
Create a dedicated test project for experimenting with new agents and prompt iterations. Compare output quality across different models before settling on a configuration.
The best agents result from many rounds of refinement. Start broad, then add specific constraints and examples based on where the output falls short.
Share your agent with a small group first. Use their feedback to improve the prompt and description before making it public on the marketplace.

​
Next Steps

Browse the Marketplace

Discover new agents and bases to add to your library

Chat Interface

Learn how to interact with agents in your projects

Project Bases

Explore templates for your next project

Creating Projects

Start a new project using items from your library