KBAISE/ for lovable
Library
Docs

Subagents

about 5 minBuildingchecked 7d agoOfficial page
The short version

Lovable can automatically launch temporary helper assistants to research, inspect code, or review work in parallel. These helpers report their findings back to the main chat without changing any of your project files directly.

Anyone planning a complex feature, investigating a bug, or exploring an unfamiliar project who wants thorough research before code changes.

Do this, in order

  1. 1

    Write a prompt asking Lovable to use subagents to inspect, research, or plan your task.

    Mentioning subagents encourages Lovable to divide the research and exploration across focused, independent helpers.

  2. 2

    Look at the activity card that appears in the project chat while the request is running.

    The card displays active subagent rows so you can see what tasks are being researched in parallel.

  3. 3

    Click on any subagent row in the activity card to see its specific discoveries.

    This reveals the exact files inspected, searches performed, and findings returned before changes are made.

  4. 4

    Review the final response or code changes applied by the main Lovable agent.

    Subagents cannot edit code, so the main agent uses their gathered findings to safely apply the final changes or summarize advice.

Paste this into your project

Use subagents to explore my project and explain how it is structured. Inspect the main pages, data flow, and key components, then summarize how everything connects before we make changes.

Words decoded

Subagent
A temporary helper assistant that Lovable starts up to investigate a specific question or inspect files.
Read-only
Able to view files, search, and browse, but completely blocked from editing, adding, or deleting any project code.

Where people get stuck

  • Expecting subagents to change your code directly; they only report findings, and all edits must come from the main Lovable agent.
  • Assuming subagents are guaranteed to run every time you ask; Lovable ultimately decides if delegating to a helper is necessary.
  • Overlooking extra credit consumption in Plan mode, where subagent research adds costs on top of the base message credit.

The short version, steps, decoder and prompt on this page are written automatically from Lovable's own documentation and can lag or misread it. The official page is always the authority.