Subagents

1. 1

   Think about a task you want Lovable to do that involves multiple steps, requires research, or needs a deep look into your project.

   Subagents are most useful for complex requests where Lovable needs to gather a lot of information before acting.

2. 2

   When you prompt Lovable, mention 'subagents' and clearly describe what specific parts of the task you want investigated.

   This encourages Lovable to use subagents for focused investigations, even though Lovable ultimately decides when to use them. Examples include 'explore my project', 'research best practices', or 'investigate a performance issue'.

3. 3

   Watch the activity card in your project chat for subagent rows to appear, showing what they are investigating and their status.

   This lets you see the progress of the investigations. You can click on a subagent row to see details of its work.

4. 4

   Review Lovable's response, which will incorporate findings from the subagents.

   Subagents report their findings to the main Lovable agent, which then uses this information to decide what to do next, continue building, or explain the results.

Use subagents to explore my project and explain how it is structured. I have not touched this app in a while and want to understand the main pages, data flow, and important files before making changes.

Subagents

These are like temporary, specialized assistants that Lovable creates to help with specific parts of a bigger job. They can look things up, inspect your project, or review work, but they can't make any changes to your files.

Main Lovable agent

This is the primary Lovable assistant you interact with directly. It's the one that receives your requests, coordinates the subagents, and ultimately makes changes to your project.

Read-only

This means something can be looked at or inspected, but it cannot be changed or edited. Subagents can only 'read' your project, not 'write' to it.

Context

This refers to all the information, background, and details that Lovable or a subagent has available to understand and complete a task. Subagents start with a fresh, focused set of information relevant to their specific investigation.

Parallel

This means multiple things are happening at the same time, rather than one after another. Lovable can run several subagents in parallel to speed up investigations.

• Expecting subagents to make changes to your project; they are currently read-only and cannot edit, create, or delete files.
• Assuming subagents automatically know everything from your chat history; they start with fresh, focused information passed to them by the main Lovable agent.
• Not explicitly mentioning subagents or the need for investigation in your prompt for complex tasks, which might lead Lovable to not use them when they could be helpful.