Feature request: persistent plan/context reference visible in UI

Resolved 💬 2 comments Opened Apr 17, 2026 by MichaelArnoldOwens Closed May 25, 2026

Problem

When working on long-running conversations that execute multi-step plans (e.g., written to a file via a planning skill), it's easy to lose track of which plan is currently being executed. This is especially painful when:

  1. The conversation is very long — the plan creation happened hundreds of messages ago and has scrolled out of view
  2. Multiple Claude Code instances are running — each working on different plans, and there's no way to glance at an instance and know which plan it's executing
  3. Resuming work after a break — you come back to a conversation and have to ask "what plan are we working on?" to re-orient

Proposed Solution

Allow pinning a reference to a file (or a short label) that stays persistently visible in the Claude Code UI — for example in the status bar, a header banner, or a collapsible sidebar element.

This could work as:

  • A pinnable file reference — e.g., /pin docs/plans/video-capabilities-plan.md makes a small indicator always visible (filename + maybe first line/title)
  • An active plan indicator — if Claude Code's plan mode is active, show which plan file is loaded
  • A status line extension — allow users or skills to set a persistent label like [Plan: video-capabilities] in the status bar

Use Case

I use a planning skill (/ce-plan) that writes structured plans to markdown files. Claude then executes the plan step by step across a long conversation. With multiple instances running different plans simultaneously, there's no quick way to identify which instance is doing what without scrolling back or asking the agent.

Even a minimal solution — like showing the plan file path in the status line — would significantly improve orientation during complex multi-plan workflows.

Alternatives Considered

  • Asking the agent — works but adds noise to every resumed conversation
  • Tasks — per-conversation and not visually prominent enough
  • Window/tab titles — could help but aren't user-configurable today
  • Memory system — persists across conversations but doesn't solve the in-conversation visibility problem

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗