[Feature Request]: Agent Concurrency & Swarm Telemetry for Main Orchestrator in Agentic ZCode

Problem Statement

In Agentic ZCode, the main orchestrator agent has the ability to dispatch, coordinate, and steer background sub-agents. However, the orchestrator lacks active visibility and telemetry into the runtime state of its sub-agent swarm:

1. Blind Orchestration: The main agent has no programmatic tool or context awareness to query how many sub-agents are currently running concurrently (e.g. distinguishing between 5, 15, or 20 active tasks).
2. Rate Limit & Concurrency Spikes: Without runtime swarm introspection, orchestrators can over-dispatch parallel tasks, accidentally exhausting account concurrency ceilings and hitting API rate limits (HTTP 429 / Error 1302: High concurrency usage).
3. Missing UI Monitoring: Users also lack an overview section indicating active vs. queued vs. finished sub-agents.

Proposed Solution / Feature Specification

1. Orchestrator Introspection Tool (manage_subagents / get_swarm_status):

   • Provide the main agent with a built-in tool or runtime context variable that reports:
     • Total active running sub-agents
     • Current concurrency vs. plan concurrency limits
     • Status of each sub-agent task (running, idle, failed, completed)
   • Allow the main orchestrator to make informed pacing and task-scheduling decisions based on live capacity.

2. User-Facing Sub-Agent Dashboard / Monitor Widget:

   • Add a visual status panel in the ZCode interface showing live sub-agent counts, remaining capacity, and execution waterfall/logs.

Community Context & Origin

• Reported by: Singh on Z.ai Community Discord
• Feedback citation:

  "The main agent can dispatch and steer agents But it doenst know how many agents are running are 20 running or 15 sub agents ? it has no clue you guys need to give the main agent orchestrator this ability especially given concurrency and rate limits"

No files, tests, or entry points are named. Start by locating the main orchestrator, sub-agent dispatch and runtime state handling, then identify the interface area for monitoring. Done should include agreed telemetry for active, queued, failed, and completed agents plus concurrency limits and a user-facing status view.