Feature Summary
Add a project-level comparison board that lets users evaluate multiple agent-created tasks side by side. The board should show task status, agent/provider, branch, diff stats, changed files, PR/check status, and quick actions to open, accept, archive, or delete a result.
Problem or Use Case
Emdash is built around running multiple coding agents in parallel, but users currently have to inspect each task one at a time. That makes it hard to compare competing solutions, identify the best result, spot stale or incomplete work, and clean up rejected attempts.
This weakens the core “parallel agents” workflow because the final decision still happens manually across separate task views.
Proposed Solution
Add a new project-level Compare view alongside Tasks, Pull Requests, and Settings.
The comparison board should include:
- A list or table of active and recently completed tasks for the project.
- Agent/provider, task name, branch name, task status, and last activity.
- Diff stats such as files changed, additions, deletions, staged/unstaged state.
- PR state when available, including draft/open/merged status and checks.
- Quick actions:
- Open task
- Open diff
- Create or open PR
- Pin/mark as preferred
- Archive rejected task
- Delete failed/unused task after confirmation
The first implementation can focus on local project/task metadata already available in the renderer stores, then expand to richer PR and test-result signals later.
Alternatives Considered
Users can currently open each task manually from the task list and compare diffs one at a time. This works for small cases but does not scale when several agents are running in parallel on the same problem.
Another option is improving the existing task list with more metadata, but that would overload the task management surface. A dedicated comparison view better matches the review-and-selection workflow.
Additional Context
Evidence from the current product surface:
- Project navigation currently has
Tasks, Pull Requests, and Settings.
- Task review capabilities already exist in individual task views: diffs, PR state, checks, conversations, and terminals.
- The missing layer is a project-level aggregation surface for comparing parallel outputs.
Feature Summary
Add a project-level comparison board that lets users evaluate multiple agent-created tasks side by side. The board should show task status, agent/provider, branch, diff stats, changed files, PR/check status, and quick actions to open, accept, archive, or delete a result.
Problem or Use Case
Emdash is built around running multiple coding agents in parallel, but users currently have to inspect each task one at a time. That makes it hard to compare competing solutions, identify the best result, spot stale or incomplete work, and clean up rejected attempts.
This weakens the core “parallel agents” workflow because the final decision still happens manually across separate task views.
Proposed Solution
Add a new project-level
Compareview alongsideTasks,Pull Requests, andSettings.The comparison board should include:
The first implementation can focus on local project/task metadata already available in the renderer stores, then expand to richer PR and test-result signals later.
Alternatives Considered
Users can currently open each task manually from the task list and compare diffs one at a time. This works for small cases but does not scale when several agents are running in parallel on the same problem.
Another option is improving the existing task list with more metadata, but that would overload the task management surface. A dedicated comparison view better matches the review-and-selection workflow.
Additional Context
Evidence from the current product surface:
Tasks,Pull Requests, andSettings.