Multi-Agent Visual Indicator for Active Sessions

Beginne damit, die Tools agent und agent_output, shell_tasks/shell_output sowie die Ereignisse des Sitzungsprotokolls nachzuverfolgen, um zu verstehen, wie der Lebenszyklusstatus des Agents für die CLI verfügbar gemacht wird. Überprüfe die vorhandenen Einstiegspunkte /status und /agents und definiere und implementiere anschließend einen kompakten Indikator, dessen Status „done“ die Anzahl der aktiven Agents, ihre Identität, eine Vorschau der Aufgabe sowie die Status „running“, „completed“, „failed“ und „queued“ abdeckt.

Feature Request: Multi-Agent Visual Indicator for Active Session

Summary

Add a visual indicator in the Command Code CLI that shows when multiple agents are running concurrently within the session, giving users real-time awareness of active subagent work — similar to how Theo implemented this for T3 Code and how Codex displays active agents.

Problem

Currently, when Command Code delegates work to multiple subagents (via the agent tool), there's no visible indicator in the UI showing:

• Which agents are currently running
• How many agents are active
• What each agent is working on (its task/description)
• Agent status (running, completed, failed)

Users are left unaware of background work happening in their session. When multiple agent calls fire in parallel (e.g., 3 explorers + 2 editors), the main session appears idle while work happens invisibly.

Proposed Solution

Implement a persistent visual indicator (status bar, footer, or inline badge) that displays:

1. Active agent count — e.g., "3 agents running"
2. Agent identity badges — show agent name/type (General, Explore, Plan, or custom names)
3. Status states — running (spinner), completed (checkmark), failed (error icon)
4. Task preview — truncated description of what each agent is doing
5. Background agent awareness — show agents started with background: true or run_in_background: true

Visual Inspiration

• T3 Code (Theo): Shows active agents as colored badges with spinners in the terminal footer
• Codex: Displays agent tiles with status indicators and task descriptions
• GitHub Copilot Workspace: Shows parallel task execution with progress indicators

User Experience

In-Session Indicator

┌─────────────────────────────────────────────────────────────┐
│  🤖 3 agents running                                        │
│  ● Explore — "Searching auth middleware..."                 │
│  ● General — "Refactoring database layer..."               │
│  ◌ Plan — "Designing API routes..." (queued)               │
└─────────────────────────────────────────────────────────────┘

Status States

• ● Running (with spinner animation)
• ✓ Completed successfully
• ✗ Failed/errored
• ◌ Queued/pending

Interaction Ideas

• Click/tap an agent badge to see its full output
• Keyboard shortcut to cycle through active agents
• /agents command enhanced to show live status alongside agent definitions
• Optional: notification when all agents complete

Technical Considerations

Existing Primitives

Command Code already has the building blocks:

• agent tool returns agent_id for background runs
• agent_output tool can check status of running agents
• shell_tasks / shell_output track background processes
• Session transcript already records agent delegation events

Implementation Approaches

1. Status bar component — persistent footer showing active agents (least intrusive)
2. Inline badges — agent indicators appear inline with conversation flow
3. Dedicated panel — toggle-able side panel with full agent details (heavier)

Agent Lifecycle Events to Surface

• Agent started (with name + task description)
• Agent progress (optional, if agents can report intermediate status)
• Agent completed (with result summary or token usage)
• Agent failed (with error reason)

Benefits

• Transparency: Users know work is happening even when main session is quiet
• Debuggability: Easier to understand what went wrong when an agent fails
• Confidence: Visual feedback that parallel delegation is working
• Discoverability: Encourages users to leverage multi-agent parallelism

Potential Concerns

• Terminal real estate: Status bar takes up space — should be compact and optionally hideable
• Noise: Could be distracting with many agents — consider collapsing/expanding
• Performance: Polling agent status shouldn't add overhead

Settings

Consider a setting to control visibility:

{
  "showAgentIndicator": true,
  "agentIndicatorStyle": "compact" // "compact" | "detailed" | "none"
}

Related Features

• Background tasks (run_in_background)
• Agent output streaming (showOutput: true)
• Session status (/status)
• /agents manager

Priority: Medium — improves UX significantly for power users leveraging multi-agent workflows

Affected Users: Anyone using parallel agent delegation, background agents, or custom subagents