[FEATURE] Shared Persistent Storage Between Claude.ai and Claude Code

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.ai and Claude Code are completely siloed. They share no context, no memory, and no data. When a user works through a complex problem conversationally in claude.ai — designing a feature, researching an API, specifying a workflow, reviewing web content — the only way to act on that work in Claude Code is for the user to manually transcribe, copy/paste, or download and transfer files. This is lossy, tedious, and breaks flow.
The permission model for scoped write access already exists in Claude Code (filesystem access with user

Proposed Solution

A persistent, user-authorized read/write space accessible from both claude.ai and Claude Code. This could be implemented as:
A shared file or directory that both tools can access
A key-value store or simple database exposed to both interfaces
An API-accessible scratchpad tied to the user's Anthropic account
The user would explicitly grant access and control permissions (read-only, read-write, or off) for each interface.

Alternative Solutions

_No response_

Priority

Critical - Blocking my work

Feature Category

CLI commands and flags

Use Case Example

1. Conversational Research → Code Action

Users frequently research and plan in claude.ai, then implement in Claude Code. Currently this requires re-explaining everything. With shared storage, claude.ai could write a structured spec, and Claude Code could read and act on it directly.

1. Web Security Logging

When claude.ai encounters prompt injection attempts or malicious content during web fetches, it currently has no way to log this persistently. With shared storage, it could write flagged URLs and details to a log that the user (or Claude Code) could process — for example, automatically generating network blocklist rules for a UniFi firewall.

1. Cross-Interface Task Handoff

A user could brainstorm and refine a task in claude.ai's conversational interface, then hand it off as a structured task to Claude Code without any manual data transfer. Think of it as a clipboard that persists across sessions and tools.

1. Data Pipeline Between Interfaces

Users with complex workflows (e.g., Shopify store management, media processing, DevOps) often need web research, API lookups, and conversational planning (claude.ai strengths) combined with filesystem operations, code execution, and automation (Claude Code strengths). Shared storage makes these complementary rather than isolated.

1. Persistent Logging and Auditing

Beyond security, users may want claude.ai to maintain running logs of decisions, recommendations, or findings across sessions — accessible to Claude Code for automation or to the user for review.

Additional Context

Why This Matters
Anthropic's product suite is converging on a model where different interfaces serve different strengths: claude.ai for conversation, research, and planning; Claude Code for execution and automation. The gap between them is currently bridged entirely by the user acting as a manual data shuttle. Closing that gap would make both products significantly more powerful and would create a workflow that no competitor currently offers.
Implementation Notes
The permission model from Claude Code (explicit user authorization for scoped access) is directly applicable
Artifact persistent storage in claude.ai already demonstrates per-user persistent key-value storage
Memory already demonstrates cross-session persistence in claude.ai
The infrastructure pieces largely exist; the missing element is cross-product access
Submitted By
A Pro subscriber who uses both claude.ai daily and Claude Code on a Linux workstation, and regularly hits the friction of transferring context between them.