Awaiting feature

[NLN6666]

Problem or motivation

To be awaited to do/run something specific commands/agents

Proposed solution

^

Alternatives considered

No response

Additional context

No response

[breaking-brake]

Thank you for submitting this feature request!

To better understand your needs, could you please provide more details?

1. Use case: What specific scenario do you want to use this "awaiting" feature for?
2. Expected behavior: What should happen when "awaiting"? For example:
   • Wait for a specific time duration?
   • Wait for an external event or condition?
   • Wait for another command/agent to complete?
3. Example workflow: If possible, could you describe a sample workflow where this feature would be useful?

This information will help us evaluate and prioritize this feature request. Thanks!

[NLN6666]

Thank you for submitting this feature request!

To better understand your needs, could you please provide more details?

1. Use case: What specific scenario do you want to use this "awaiting" feature for?

2. Expected behavior: What should happen when "awaiting"? For example:

   • Wait for a specific time duration?
   • Wait for an external event or condition?
   • Wait for another command/agent to complete?

3. Example workflow: If possible, could you describe a sample workflow where this feature would be useful?

This information will help us evaluate and prioritize this feature request. Thanks!

It is similar to Claude Code's Sub Agents. Its function is to batch execute multiple commands (as mentioned in my other issue) or Agents at once. The parameter is the waiting time or waiting for completion (waiting for another sub agent to complete its task or executing immediately after the {{VAR}} variable is satisfied) I don't know if this will be difficult to implement (maybe Claude Code doesn't support it or something) It is similar to C++'s mutex asynchronous thread Forgive me for not being very familiar with these, I am still a beginner

[breaking-brake]

Thank you for your patience.

I'm currently investigating the feasibility of this feature. Please bear with me.

[breaking-brake]

Thank you for waiting. The functionality you're looking for is already achievable with the current features.

Parallel Execution & Await

How it works

You can use SubAgent or SubAgentFlow nodes for parallel execution. Claude Code CLI automatically handles the synchronization (await).

                         ┌─────────────────────┐
                         │       Start         │
                         └──────────┬──────────┘
                                    │
              ┌─────────────────────┼─────────────────────┐
              │                     │                     │
              ▼                     ▼                     ▼
   ┌──────────────────┐  ┌──────────────────┐  ┌──────────────────┐
   │  SubAgent Node   │  │  SubAgent Node   │  │  SubAgent Node   │
   │  (Security Check)│  │ (Performance)    │  │  (Style Check)   │
   └────────┬─────────┘  └────────┬─────────┘  └────────┬─────────┘
            │                     │                     │
            │        ┌────────────┴────────────┐        │
            │        │   Automatic Await by    │        │
            └───────▶│   Claude Code CLI       │◀───────┘
                     └────────────┬────────────┘
                                  │
                                  ▼
                         ┌─────────────────────┐
                         │        End          │
                         └─────────────────────┘

CLI Output Example

When SubAgents run in parallel, Claude Code CLI shows:

⏺ 3 agents finished (ctrl+o to expand)
   ├─ security-check-subagent · 24 tool uses · 73.7k tokens
   ├─ performance-check-subagent · 18 tool uses · 66.9k tokens
   └─ style-check-subagent · 20 tool uses · 49.6k tokens

For Skills, MCP & Prompt Parallel Execution

If you want to run Skills, MCP tools, or Prompts in parallel, use them inside a SubAgentFlow:

   ┌──────────────────────────────────────────────────────────┐
   │                     SubAgentFlow                         │
   │  ┌─────────┐    ┌─────────┐    ┌─────────┐              │
   │  │ Prompt  │    │  Skill  │    │   MCP   │              │
   │  │  Node   │───▶│  Node   │───▶│  Node   │              │
   │  └─────────┘    └─────────┘    └─────────┘              │
   └──────────────────────────────────────────────────────────┘

Multiple SubAgentFlows can run in parallel, achieving parallel execution of Prompts/Skills/MCP.

---

Does this cover your use case? Let me know if you have any questions.

[NLN6666]

Oh didnt know Subagent Flow this option. You idea is great. Thanks for the reply!

[NLN6666]

In my version(latest), subagent flow doesnt support run multi-agents?

[NLN6666]

在我最新版本中，子代理流程不支持运行多代理？

oh sorry it indeed doenst support can you make this a feature or just a Prompt and ask to run a subagent?

[breaking-brake]

Hi @NLN6666,

Thanks for the follow-up!

Actually, SubAgentFlow parallel execution is already supported and working in my environment.
Here's an example from my testing (console log):

⏺ Running 2 agents… (ctrl+o to expand)
   ├─ my-workflow_subagentflow-1 (Sub-Agent Flow: subagentflow-1) · 0 tool uses · 40.4k tokens
   │  ⎿  Initializing…
   └─ my-workflow_subagentflow-2 (Sub-Agent Flow: subagentflow-2) · 1 tool use · 40.9k tokens
      ⎿  Done

As you can see, two SubAgentFlow nodes (subagentflow-1 and subagentflow-2) are running in parallel.

I may not fully understand your use case yet.
If possible, could you also share:

• Your workflow JSON file (from .vscode/workflows/xxx.json)
• A screenshot of the console output when you ran the workflow

This will help me understand whether there's a bug or if there's something specific about your workflow configuration.

[NLN6666]

I mean in a subagent flow you can run subagents in it but its not very necessary I think, it has other ways to implement it

[NLN6666]

General-worldflow.jsonThis is my workflow, perhaps you need to translate it to English before reading it

[breaking-brake]

Hi @NLN6666,

Thank you for waiting.

Looking at your workflow, I noticed you have a "parallel-worker" SubAgent designed to spawn and coordinate other sub-agents. Let me address a few points:

Parallel Execution at Top Level

Although this may not be your main question, just to clarify: if you want to run multiple SubAgents in parallel, you can connect the previous node directly to multiple SubAgent nodes without using a Switch node. Switch only selects one branch, so it prevents parallel execution.

Nested SubAgents (SubAgent within SubAgent)

Unfortunately, running SubAgents within a SubAgent is not currently possible due to Claude Code CLI limitations. The CLI does not support spawning nested Task agents from within a Task agent.

Possible Alternative: Skills

If you want to achieve similar behavior, Skills might be an option. You can use Skill nodes within a SubAgentFlow to create modular, reusable logic, though it's not true nested parallelism.

---

If this doesn't address your concern, I'd like to move this conversation to GitHub Discussions for a more open-ended exploration of your use case. Please let me know!

[NLN6666]

Yes, recently I have indeed been converting my workflow to be a skill. And thanks for your clarification and correction! I think now I've got all my problems fixed, so you can close this issue now. Thanks again for replying!

[breaking-brake]

Thank you for your understanding! I hope your Skill creation goes smoothly.

By the way, we're planning to enforce lowercase-only workflow names in an upcoming release (see #415). Your workflow General-worldflow will need to be renamed to general-worldflow after the update. Just a heads-up!