Problem
IDE-first users expect an assistant to understand the file/range/diagnostic they are looking at. CodeWhale should support that job without becoming a full IDE fork. The terminal/TUI should remain the source of truth, while editors can hand it precise context.
Scope
For v0.8.47, design and implement the first editor-context bridge around ordinary CodeWhale sessions.
- Start with VS Code as the common denominator for classrooms and contributors.
- Support sending selected files/ranges, current diagnostics, and git diffs into the active CodeWhale session.
- Prefer protocol bridges such as ACP, MCP, or the local runtime API before creating a large first-party IDE surface.
- Keep the terminal/TUI session as the canonical session record.
- Do not attempt full inline code completion until workspace-indexing and latency are strong enough.
- Keep auth and runtime-token handling explicit.
Acceptance criteria
- A VS Code user can send current editor context to CodeWhale without copy/paste.
- The receiving CodeWhale session shows what context was added and from where.
- The bridge does not silently expose the runtime API beyond localhost/private configuration.
- Docs explain the difference between editor context handoff and a full IDE plugin.
Related: #1264, #1584, #1801
Related tracker: #1881
Problem
IDE-first users expect an assistant to understand the file/range/diagnostic they are looking at. CodeWhale should support that job without becoming a full IDE fork. The terminal/TUI should remain the source of truth, while editors can hand it precise context.
Scope
For v0.8.47, design and implement the first editor-context bridge around ordinary CodeWhale sessions.
Acceptance criteria
Related: #1264, #1584, #1801
Related tracker: #1881