claude-code - 💡(How to fix) Fix [FEATURE] "Send to Claude Code" action from a Project chat (analogous to the Figma Design → Claude Code handoff) [1 comments, 2 participants]

Preflight Checklist

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

Problem Statement

When working inside a Project in Claude Desktop, my workflow today is:

Open the project. Ask Claude something — usually clarification, scoping, or planning that benefits from the project's accumulated context, custom instructions, and attached files. Manually copy Claude's resulting prompt / plan / spec. Paste it into Claude Code in a separate window/terminal so it can actually execute against my codebase.

That last copy-paste hop is friction, and it loses structure. If the assistant produced a prompt that includes file references, code blocks, or follow-up steps, I have to manually reconstruct that as a single coherent input for Claude Code. The hand-off is also lossy w.r.t. project context — Claude Code doesn't automatically inherit anything I set up in the Project. There is already a precedent for solving exactly this in another integration: from a Figma Design surface, there is an "Open in Claude Code" action that ships the design context directly into a Claude Code session in one click. Doing the same from a Project chat would be the natural extension.

Proposed Solution

Add a "Send to Claude Code" action available from a Project chat (and ideally from any chat). Two reasonable placements:

A button in the chat composer area (next to the model picker / + attachment button), so the user can decide before sending: "this whole prompt should go to Claude Code, not be answered here." A per-message action on any assistant response (similar to the existing copy / regenerate actions), so the user can take a prompt Claude just produced and forward it to Claude Code in one click.

What gets handed off:

The selected prompt or assistant message body (preserved as-is, including markdown, fenced code blocks, and file references). Optionally, a compact summary of the project context (custom instructions, list of attached files / project knowledge) so Claude Code starts with the same baseline the project chat had. A way for the user to pick which Claude Code session/working directory to deliver into, with the most recently used one as the default.

Alternative Solutions

Status quo (manual copy-paste). Works, but it's the friction this request exists to remove, and it loses formatting fidelity. MCP-based bridge. Setting up a custom MCP server to forward content from Desktop to Code is overkill for a workflow that's already supported in spirit by the Figma → Code path. Just opening the same prompt in Code mode of the same Desktop window. Not equivalent — Claude Code typically runs against a specific working directory and has tools (Bash, file I/O, MCPs scoped to the repo) that the chat surface does not.

Priority

Critical - Blocking my work

Feature Category

CLI commands and flags

Use Case Example

No response

Additional Context

The workflow this targets — "think in chat with full project context, then execute in Claude Code" — is, anecdotally, already how a lot of users mix the two surfaces. Making it first-class would (a) reduce manual steps, (b) reduce errors from imperfect copy-paste, and (c) put structured context (file refs, code blocks) into Claude Code in a form it can act on directly. Same direction the Figma integration already established — this would just generalize that affordance to plain text/markdown prompts originating in a Project chat.

Claude for Mac version: 1.4758.0 (fb266c) macOS: Tahoe 26.4.1