Lovable Plan mode: choose a plan before changing code

Lovable Plan mode investigates a change and creates a structured plan you can edit before approving implementation. Select Plan in the mode picker beside the chat input; selecting Approve on the finished plan switches to Build and starts the work, according to Lovable’s Plan guide↗ (opens in a new tab). Tools like fenbs keep the decision and unfinished checks outside that conversation, ready when you return or use another AI.

Which mode should you choose?

The distinction is the result you want next. Chat helps you understand the app or explore an idea without changing its code or data. Plan investigates and produces an editable plan. Build implements changes. The conversation carries across these modes, so discussing an idea does not require starting over when you decide to plan it. Lovable explains these boundaries in its Chat mode guide↗ (opens in a new tab).

For example, suppose you want to reduce a long signup form. Start by asking which fields are essential. Once you have chosen the smaller form, request a plan that names the affected screens and preserves required information. A vague request to make signup better leaves too much room for a change you did not mean.

How do you start and review a plan?

Use the mode picker next to the chat input and choose Plan. Lovable also documents Option+P on Mac and Alt+P on Windows or Linux to cycle modes. In Plan view you can review and edit the proposal before selecting Approve. Skip declines the plan without starting that build; the Plan instructions↗ (opens in a new tab) describe the complete flow.

1. Describe one outcome in a sentence. Name the screen, behavior, or error you want to change.
2. List the things that must keep working. Include any decision you have already made, such as keeping the current sign-in method.
3. Ask for a plan with the affected parts, the order of work, and a way to check each outcome.
4. Read the proposal as a set of commitments. Edit anything that adds unrelated work or assumes an answer you have not given.
5. Approve only when the plan matches your intended change and you are ready for implementation to start.

A good review catches missing boundaries. If the request concerns the signup form, ask whether the plan also changes account creation or existing accounts. A plan that names the visible screen but ignores what happens after submission is not complete enough to approve. You should be able to explain the expected outcome without quoting the chat.

What should your planning request include?

Give the investigation a narrow question and a clear stopping point. You do not need to prescribe every file. You do need to distinguish a requirement from a suggestion, and a confirmed problem from your guess about its cause. That helps you judge whether the proposed implementation solves the original need.

Goal: shorten the signup form on phones.
Keep: the current sign-in method and existing account behavior.
Investigate: which fields are required and where their values are used.
Plan: the smallest change, the affected screens, and the checks.
Questions: list anything that needs my decision before approval.

Add a small example of success: a person can finish the form with one hand, see a clear error for a missing required field, and reach the expected next screen. These are review criteria you choose, rather than a promise about what an AI will deliver. For more starting requests, see the Lovable prompt library.

Save the reason for the chosen approach as well as the steps. A later assistant might otherwise reopen a trade-off you already settled. The AI at work guides cover keeping useful outcomes from a session, while how fenbs works explains how connected assistants read the same decisions before continuing.

What are the important limits before you approve?

Plan is a preparation stage, not a promise of free work. Lovable documents planning usage and separately billed Build implementation. Check its current credits and usage guidance↗ (opens in a new tab) before a broad investigation or a series of revisions. Keep the question specific instead of asking for an unrestricted review of everything.

Chat’s read-only boundary applies to the app’s code and data. Connected tools are different: Lovable can perform writes through them after approval, or under a saved Always allow choice. Review tool requests even when you are only discussing a change. This distinction matters if a conversation can create work or send something to another service.

Approval is also a change in activity. It starts implementation rather than simply filing the plan for later. Before you select it, check that the scope is current, that open questions have answers, and that your acceptance checks describe behavior you can actually inspect. If a plan feels too large to review, split the desired result into smaller changes.

What should you keep after planning and building?

Keep four short things: the chosen approach, why you chose it, what was implemented, and what still needs checking. Separate an approved plan from a completed result. A future reader should not have to infer whether a step happened merely because it appears in a well-written proposal.

Lovable stores the active plan in .lovable/plan.md and archives approved plans in dated files under .lovable/plan/, as its planning guide↗ (opens in a new tab) explains. Those files preserve the proposal. Add your own concise outcome beside the work so a later session can tell which checks passed and which decisions changed.

Keep neighboring questions separate. The custom-domain checklist handles publication checks, and Lovable versus Claude Code covers choosing a working approach across tools. Neither replaces reviewing this particular change before implementation.

How to keep your approved plan with fenbs

Keep the work so you can carry on from exactly where you left off, on any device and in any other AI. Save the small record someone needs to continue, including what you approved and what remains untested.

1. Create a free account.
2. Connect Lovable to fenbs through Integrations. Its chat connector is personal to you and may be restricted by your workspace administrator.
3. Save the change you want, the plan you approved, and the reason for your choice in plain words.
4. After the build, ask your assistant to save what changed, what you checked, and the first unfinished step.
5. When you return, ask your connected AI to read that record before doing more work.

Lovable documents personal connections and administrator controls in its custom connector guide↗ (opens in a new tab). fenbs keeps the decisions and progress; it does not inspect or implement the plan for you. For one small change you can finish and check in a sitting, a short local note may be enough.