Description
Problem
When I start agents from issues (Tasks → Start, or Create worktree with a linked issue), Orca launches the agent with a first prompt that already contains the issue number or link. The agent starts working immediately, so I can't run /model first. The model is fixed by the provider's single Arguments field.
I often want to start several agents from a list of issues and choose the model per issue. For example, Opus for the hard ones and Sonnet for the routine ones. Today I have to edit Settings → Agents → Arguments between launches, which is slow and easy to forget, and not workable when starting several in a row.
Proposed behavior
- Let users create additional agent entries for a provider ("Duplicate" or "Add preset"). Each one has:
- a name (e.g. "Claude Opus")
- its own Command, Arguments and environment
- its own Enabled/Disabled and Default state
- Show the presets in the Agent dropdown of Create worktree, and wherever else an agent is picked when starting from an issue or PR. Group them under the provider's icon.
- Remember the last-used preset per project, or let the user set a default.
Example
| Preset | Command | Arguments |
|---|---|---|
| Claude Opus | claude |
--model opus |
| Claude Sonnet | claude |
--model sonnet |
Why this matters
- Choosing a model per issue is a common workflow when launching agents in bulk. Opus fits planning and hard problems, Sonnet fits cheaper, faster tasks.
- The first prompt is sent at launch, so the model has to be chosen before the session starts.
/modelcan't fix that. - "Claude Agent Teams" already shows Orca supports more than one launch configuration per provider. Users just can't define their own.
Alternatives considered
- A per-launch override: the sliders icon next to "Agent" in Create worktree could accept one-off arguments. Helpful, but saved presets are faster when launching many agents.
- A wrapper script per variant also needs the ability to add more than one entry.
Screenshots
(Settings → Agents with a single Arguments field per provider, and the Create worktree dialog's Agent dropdown.)
Environment: Orca on Linux (Omarchy), Claude CLI.
Activity
cfqlilei commentedon Oct 31, 2026
Concrete use case: named Claude launch variants for different models.
I use multiple Claude Code model configurations and want to select them directly when creating or launching an agent, for example:
Claude Sonnet→claude --model sonnetClaude Opus→claude --model opusThe desired behavior is:
To keep the initial UX simple, the main requirement is reusable Claude
--modelarguments. Separate controls for effort, permissions, or account management are not required for this use case.Environment: Orca 1.4.245 on macOS 15.5 (Apple Silicon), Claude Code 2.1.311.
TomaszMarciniakHeroGaming commentedon Nov 12, 2026
+1. My main use is switching between two Claude models (Opus / Fable) when I create a worktree. Showing presets as one-click options in Create worktree, not only in the dropdown, would remove most of the friction.