anthropics / anthropics/claude-code
[FEATURE] Deep link to resume an existing Cowork session (claude://cowork/{session-id})
- Dominant language
- Python
- Stars
- 145k
- Forks
- 23.1k
- PR merge metrics
- PR metrics pending
Description
## Summary
Claude Desktop's documented deep links (https://support.claude.com/en/articles/14729294-open-claude-desktop-with-a-link) cover new chats, existing claude.ai chats/projects, and **new** Cowork/Code sessions — but there is no link that resumes an **existing local Cowork session** by ID.
## Ask
```
claude://cowork/{session-id}
```
Opens Claude Desktop directly to that Cowork session (or falls back to the session list for an unknown ID — same behavior `claude://claude.ai/chat/{id}` already has for invalid IDs).
## Why it should be cheap to add
- The `claude://` protocol handler and its router already exist and already resolve IDs for `claude.ai/chat/{id}`.
- Session IDs are already stable and known: `~/Library/Application Support/Claude/local-agent-mode-sessions///local_.json`.
- The app already navigates to these sessions from its own sidebar; the deep link only needs to trigger that same navigation.
## Use case
External launchers (Stream Deck, Raycast, scripts, agent dashboards) that monitor multiple Cowork sessions and want to jump to the one that needs attention. Today the only path is: open app, find session in sidebar, click.
## Related
- #50345 — same ask (resume by UUID), closed as duplicate
- #54614 — new Code session in folder via deep link
- #43943 — `claude-desktop://resume?session=`
## Platform
macOS, Claude Desktop (Cowork research preview).
Contributor guide
No contributing guide indexed for this repository
Research direction
Start by locating the existing claude:// protocol handler and router, especially the route that resolves claude.ai/chat/{id}. Trace how the app navigates to local Cowork sessions from its sidebar, then verify that claude://cowork/{session-id} opens a matching session and falls back to the session list for an unknown ID.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- macos
- Domain
- desktop
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 38/100