When to Use Plan Mode
Trace has three AI interaction modes. Choosing the right one makes a real difference:
| Mode | Best For |
|---|
| Ask | Quick questions, design explanations, “how does this circuit work?” |
| Agent | Immediate edits — add a capacitor, route a trace, generate Gerbers |
| Plan | Complex multi-step tasks — design a power supply, lay out an entire board, refactor a schematic |
The Plan Mode Lifecycle
Plan mode follows a structured lifecycle. Each phase gives you a checkpoint before the AI proceeds.
1. Research Phase
When you submit a request in Plan mode, the AI first researches your task before proposing anything. Research runs server-side and may include:
- Web search — Finding datasheets, reference designs, application notes, and technical documentation
- Parts search — Querying component databases via Nexar/Octopart, DigiKey, and Mouser for availability, pricing, and specs
- Datasheet parsing — Extracting pinouts, recommended circuits, and electrical parameters from PDFs
- Symbol and footprint search — Finding matching components in the library via vector similarity
A Research Summary Card appears in the chat as a collapsible panel. It streams in real-time so you can see what the AI is finding as it works. The card summarizes key findings in 2-4 sentences once research completes.
The research phase uses a multi-agent orchestrator that runs multiple searches in parallel. This means the AI can simultaneously search for components, read datasheets, and look up reference designs — significantly faster than sequential research.
2. Clarification Questions
If the AI needs more information before it can create a solid plan, it presents structured questions in a PlanQuestionsBar at the bottom of the chat. These are typically multiple-choice questions about your requirements:
- Target voltage and current specs
- Preferred component packages (SMD vs through-hole)
- Board size constraints
- Layer count preferences
- Budget or sourcing preferences
Answer the questions and the AI incorporates your responses into the plan. The AI may ask follow-up questions if your answers raise new considerations.
3. Plan Document
Once research and questions are complete, the AI generates a structured plan document that includes:
- Architecture overview — A Mermaid diagram showing the system block diagram, signal flow, or circuit topology
- Component table — Selected parts with values, packages, and distributor links
- Implementation checklist — Numbered steps the AI will execute if you approve
- Design rationale — Why specific components or approaches were chosen
The plan document renders inline in the chat with full Markdown formatting, including Mermaid diagrams that render as interactive SVGs.
4. Approval Workflow
After reviewing the plan, you have four options:
| Action | What Happens |
|---|
| Approve | The AI begins executing all steps. A todo checklist is created from the implementation steps and tracks progress in real-time. |
| Modify | Tell the AI what to change. It adjusts the plan and presents it again for review. |
| Reject | Cancel the plan entirely. No changes are made to your design. |
| More Research | Ask the AI to investigate further — different components, alternative approaches, additional reference designs — before finalizing. |
5. Execution with Todo Tracking
When you approve a plan, the AI creates a todo checklist from the implementation steps. As it works through each step, the checklist updates in real-time:
- Each step shows as pending, in-progress, or completed
- The todo widget appears in the chat panel and expands to show all items
- If the AI encounters an issue on a step, it explains the problem and adjusts its approach
- You can see exactly where the AI is in the process at any time
Extended Thinking
During all phases of Plan mode (and in Agent mode for complex tasks), the AI shows its reasoning process in real-time via collapsible thinking blocks. These let you see:
- How the AI is analyzing your design
- What tradeoffs it’s considering
- Why it chose a particular approach
- What constraints it’s working around
Thinking blocks appear inline in the activity feed and auto-collapse when the AI moves to the next step. Click to expand and read the full reasoning at any time.
Extended thinking uses a 10,000-token budget per reasoning step, allowing deep analysis of complex design decisions.
Tips for Effective Plan Mode Usage
- Be specific about constraints — “Design a 5V 3A buck converter using available JLCPCB parts in a SOT-23-5 package” gives the AI much more to work with than “design a power supply.”
- Attach datasheets — If you have a specific component in mind, drag the datasheet into the chat. The AI will extract the recommended circuit directly.
- Use the questions phase — Don’t skip past questions. They help the AI avoid assumptions that lead to rework.
- Review the Mermaid diagram — The block diagram in the plan document is the fastest way to verify the AI understood your architecture.
- Modify rather than reject — If the plan is 80% right, modify it. The AI preserves what’s good and adjusts what you flag.
