Feature request: access to claude.ai/app chat transcripts and more durable memory

Problem

Claude Code has search_session_transcripts and list_events which work well for retrieving context from Claude Code sessions. However, many users do significant work in claude.ai (web) or the mobile app before moving to Claude Code — and those transcripts are completely inaccessible.

In my case, I built an elaborate Excel workbook across 8 versions over multiple claude.ai chat sessions (extracting lab results from 3 hospital portals, building interactive dashboards, running adversarial audits with 8 agents). When I later needed that context in Claude Code, the only way to recover it was to manually copy-paste the entire chat history into a .docx file and have Claude Code read that.

This is a real friction point: the AI that helped me build something can't remember having built it.

Suggestion 1: Cross-platform transcript access

Allow Claude Code to search/read transcripts from claude.ai web and mobile app conversations — not just Claude Code sessions. These could be:

• Indexed locally when the user syncs (opt-in)
• Accessible via API if the user authorizes
• Imported from exported chat archives

Suggestion 2: More durable, automatic memory

The current memory system (manually saving .md files to a project directory) works but is fragile — it depends on the AI deciding to save, and the user remembering to ask. A mechanism that automatically captured key artifacts, decisions, and development history across sessions (regardless of client — web, mobile, CLI) would be far more robust.

Think of it as a project journal that writes itself: what was built, what changed between versions, what decisions were made and why, what was tried and rejected.

Context

I'm a power user running Claude Code on Windows 11, and I use claude.ai web, the desktop app, and Claude Code interchangeably depending on the task. The lack of cross-platform memory continuity is the single biggest source of friction in my workflow.

---
Filed via Claude Code on behalf of the user.