Listen to today's AI briefing

Daily podcast — 5 min, AI-narrated summary of top stories

Claude Code's /ultraplan Command Offloads Complex Planning to the Cloud

Claude Code's /ultraplan Command Offloads Complex Planning to the Cloud

Ultraplan is a new research preview feature that generates complex coding plans remotely, allowing for targeted feedback and flexible execution either on the web or back in your terminal.

GAla Smith & AI Research Desk·1d ago·3 min read·1 views·AI-Generated
Share:
Source: code.claude.comvia hn_claude_codeSingle Source
Claude Code's /ultraplan Command Offloads Complex Planning to the Cloud

What Changed — A New Cloud-Powered Planning Mode

Claude Code has introduced a new feature called Ultraplan, currently in research preview. This is not just another planning mode—it's a hybrid workflow that generates detailed implementation plans in a dedicated cloud session on claude.ai, separate from your local terminal. This follows a broader trend of Claude Code expanding its cloud capabilities, as seen with the recent launch of Claude Code on the web and the Computer Use feature in late March.

What It Means For You — Terminal Freedom and Collaborative Review

The core value is asynchronous, hands-off planning. When you trigger Ultraplan, Claude generates the plan remotely. Your terminal shows a status indicator (◆ ultraplan ready), but remains completely free for other commands or sessions. This solves the problem of your main coding session being blocked during long, complex planning phases.

More importantly, it enables a collaborative review workflow. You open the plan in your browser on claude.ai, where you get:

  • Inline comments: Highlight any passage and leave specific feedback for Claude to address.
  • Emoji reactions: Quickly signal approval or concern on sections.
  • Outline sidebar: Navigate large plans efficiently.

This targeted feedback loop is a significant upgrade from replying to an entire plan in your terminal, where context can get messy.

How To Launch It — Three Simple Triggers

From your local CLI, launch Ultraplan in three ways:

  1. Direct command: /ultraplan <your prompt>
  2. Keyword trigger: Include the word ultraplan anywhere in a normal prompt.
  3. From a local plan: When Claude finishes a local plan and shows the approval dialog, choose "No, refine with Ultraplan on Claude Code on the web".

To manage the session, run /tasks and select the ultraplan entry. You can stop it at any time, which archives the cloud session and clears the terminal indicator.

The Critical Choice — Where To Execute

Once the plan is finalized in the browser, you choose the execution path:

Option 1: Execute on the web
Select "Approve Claude’s plan and start coding in your browser." Claude implements the plan in the same cloud session. When finished, you review the diff and can create a pull request directly from the web interface. Your terminal gets a confirmation and the status clears.

Option 2: Send back to terminal
Select "Approve plan and teleport back to terminal." This option appears only if your CLI session is still active and polling. The web session archives, and your terminal shows a dialog titled "Ultraplan approved" with three choices:

  • Implement here: Inject the plan into your current conversation and continue.
  • Start new session: Clear current conversation and begin fresh with only the plan as context.
  • Cancel: Save the plan to a file. Claude prints the file path for later use with claude --resume.

When To Use Ultraplan vs. Local Planning

Use Ultraplan when:

  • The task requires multi-step, complex reasoning that would block your terminal.
  • You want to collaborate with others or provide structured feedback on specific plan sections.
  • You're considering a web-based PR workflow for the implementation.

Use local plan mode when:

  • The task is straightforward and you want immediate, iterative execution.
  • You need full access to your local environment during the planning phase itself.
  • You're working offline or prefer to keep everything in-terminal.

This new hybrid approach gives you a strategic advantage for architecting larger features or refactors without sacrificing terminal availability for other work.

Following this story?

Get a weekly digest with AI predictions, trends, and analysis — free.

AI Analysis

**Immediate Action:** Start using `/ultraplan` for any feature spec or refactor that feels too large for a single terminal session. The ability to comment on specific sections is a game-changer for precision. Instead of saying "redo the database layer," you can highlight exactly which part of the plan is problematic. **Workflow Shift:** Treat your terminal as the execution engine and the web interface as the planning whiteboard. Kick off Ultraplan at the start of a complex task, then switch to other terminal work (like running tests on a different branch) while Claude thinks. This turns planning from a blocking operation into a parallel process. **Integration Point:** This feature directly complements the **Conductor MCP** server we covered on April 4th, which orchestrates multiple Claude Code sessions. You could use Conductor to manage your main terminal session while an Ultraplan generates in the background, effectively multi-tasking with Claude agents.
Enjoyed this article?
Share:

Related Articles

More in Products & Launches

View all