[FEATURE] Pinned, editable plan and checklist that stays visible while Claude works

Status Open
Maintainer reply None cached
Activity 0 comments · opened Oct 11, 2026

Preflight Checklist

  • [x] I have searched existing requests and this feature hasn't been requested yet
  • [x] This is a single feature request (not multiple features)

Problem Statement

Claude Code already makes plans (plan mode) and keeps a todo list while it works, but both scroll away inside the conversation:

  • The plan disappears. After approving a plan, it's buried under tool output within minutes. To check "where are we against the plan?" I scroll back or ask Claude.
  • The checklist moves. The todo list is redrawn in the stream each time it changes. In long turns I can't glance at progress.
  • I can't edit either one. If I want to reorder steps, drop one, or add "also update the docs", I have to type it as a chat message and hope it's applied to the list.

Proposed Solution

A pinned plan panel at the top of the chat (in the IDE extensions) or a collapsible header (in the terminal), shown whenever a plan or todo list exists:

┌ Plan: Migrate auth to new token API ─────────── 3/6 ┐
│ ✅ 1. Add TokenClient wrapper                        │
│ ✅ 2. Replace calls in session.ts                    │
│ ✅ 3. Update login flow                              │
│ ▶  4. Fix refresh-token tests          (in progress) │
│ ○  5. Remove old client                              │
│ ○  6. Update docs                       [added by me]│
└──────────────────────────────────────── [Edit plan] ┘
  1. Always visible. It stays pinned while output scrolls, with progress such as 3/6 and the current step highlighted.
  2. One source of truth. The approved plan becomes the checklist, so the plan and the todos don't drift apart.
  3. Editable by me. I can tick, untick, reorder, add or delete items, and edit the plan text. Claude receives the change as context on its next step, for example "user added step 6, removed step 5".
  4. Click to jump. Clicking a step jumps to the part of the transcript where it was worked on.
  5. Survives compaction and /resume. The plan is session state, not just a message in the history.

Alternative Solutions

  • Scrolling back, or asking "what's left?": slow, and it costs tokens.
  • Keeping the plan in a markdown file in the repo: it works, but it's manual and doesn't update live.
  • Other tools: Cursor shows the agent's todo list pinned above the chat input, with progress, and its plan mode produces an editable plan document. That workflow makes long agent tasks much easier to supervise.
  • Related, closed: #76537 (sticky todo list while scrolling) and #66453 (a single plan pane combining plan mode, tasks and specs). This request combines the two and adds two-way editing.

Priority

High - Significant impact on productivity

Feature Category

Interactive mode (TUI)

Use Case Example

  1. I approve a 6-step plan and switch to reviewing a PR in another window.
  2. Glancing back, the pinned panel shows 4/6, fixing refresh-token tests, and I don't need to read the output.
  3. I realize the docs need updating, so I add step 6 directly in the panel. Claude picks it up when it reaches that point.
  4. The next morning I /resume, and the panel shows exactly where we stopped.

View original on GitHub ↗