Feature: Label-based Runner Selection (Claude vs Codex vs Cursor)

[hugolrf]

Feature Request

Problem

Currently, label-based routing supports:

• Model selection (opus, sonnet, gemini-*)
• Mode selection (debugger, builder, scoper, orchestrator)
• Tool restrictions

But there's no way to route issues to different AI providers/runners (Claude Code vs Codex vs Cursor) based on Linear labels.

Use Case

Users want to:

• Route quick fixes to @cursor label → uses Cursor
• Route complex tasks to @claude label → uses Claude Code
• Route specific code tasks to @codex label → uses Codex

Proposed Solution

Add runner to labelPrompts configuration:

{
  "labelPrompts": {
    "debugger": {
      "labels": ["Bug"],
      "runner": "claude"
    },
    "fast-fix": {
      "labels": ["QuickFix"],
      "runner": "cursor"
    },
    "code-review": {
      "labels": ["CodeReview"],
      "runner": "codex"
    }
  }
}

Or via runner-specific labels:

{
  "runnerLabels": {
    "claude": ["@claude", "complex"],
    "codex": ["@codex", "refactor"],
    "cursor": ["@cursor", "quick-fix"]
  }
}

Alternatives Considered

1. Multiple Cyrus instances - each with different defaultRunner - adds complexity
2. Manual assignment in Linear - requires manual work

Environment

• Cyrus version: latest
• Installation: self-hosted
• Platform: macOS/Linux

[hugolrf]

Context: Self-hosted CLI installation (not cloud). Users running multiple AI CLI tools locally (Claude Code, Codex, Cursor) want to route issues to specific providers based on Linear labels.

Currently RunnerType enum exists (["claude", "gemini", "codex", "cursor"]) but no label-based selection mechanism.

[Connoropolous]

as a quick interim solution you could use

agent=codex

in the issue description