claude-code - 💡(How to fix) Fix [FEATURE] In-product feedback-to-remedy loop for Max plan users

Problem

When Claude makes errors that consume significant token budget (especially errors about its own product features), Max plan users have no efficient path to resolution. Current options:

• /feedback — no ticket tracking, no response guarantee, no remedy mechanism
• support.anthropic.com — context switch, re-explain everything, no session linking
• GitHub Issues — community channel, not a billing support path

Competitors (GitHub Copilot) have integrated error reporting that connects session context to support infrastructure in one click.

Proposed Solution

A built-in feedback-to-remedy workflow:

1. User flags a session issue via /feedback or a UI button
2. System auto-attaches: session ID, token usage, transcript excerpt, timestamp, plan type
3. Creates a trackable ticket (visible to user with status updates)
4. For token-waste issues on paid plans: support can apply credit without the user leaving the tool
5. User receives resolution notification in-product

Why This Matters

• Anthropic already has the telemetry data for every session
• Anthropic already has the billing system for Max plans
• The only missing piece is the glue between them
• Trust is critical for paid plan retention — users need to know that when Claude burns their budget on its own errors, there is a real recourse path
• The current experience is: error happens → user loses tokens → user must leave the product to file a report → no tracking → no guarantee of resolution

Real-World Example

Session 71b8d81b-1505-46d8-afe3-d84b74baad64 (2026-05-29): ~2M tokens consumed due to Claude providing incorrect guidance about its own /agents feature. The user had no efficient way to flag this for remedy without leaving the product, re-explaining the situation on a web form, and hoping someone reads it.

See: https://github.com/anthropics/claude-code/issues/63554