Sub-agents

Complex changes often involve independent pieces of work — updating copy in one section while restructuring layout in another. Instead of doing everything sequentially, Coframe Agent‘s agent can delegate tasks to sub-agents that work in parallel.

How delegation works

1. The agent analyzes your request and identifies tasks that can run independently.
2. It spawns sub-agents, each assigned a specific objective — “Update the hero section copy” or “Add responsive breakpoints to the pricing grid”.
3. Sub-agents run concurrently, each with their own:
   • Chat transcript — their reasoning and progress.
   • Task plan — their breakdown of the assigned objective.
   • Tool access — they can edit files, read the workspace, and verify their work.
4. Results flow back to the parent agent, which integrates them into the final change.

Full scenario walkthrough

Here’s the complete sub-agent flow — from the initial user request through delegation, monitoring, and drilling into a specific agent’s work.

1. Chat — user request triggers delegation

The user describes a change that involves independent tasks. The agent recognizes this and delegates to sub-agents automatically:

2. Sub-agents list — parallel agents at work

The Sub-agents panel shows each delegated agent with its name, objective, status badge, and elapsed time. You can see at a glance which agents are still running and which have completed:

3. Agent detail — drill into a sub-agent’s transcript

Click any sub-agent to see its full chat transcript and reasoning. The detail view uses the same components as the main chat — tool calls, code diffs, and all:

The Sub-agents panel

The Sub-agents panel shows every delegated agent in real time:

• Agent list — name, objective, status, and elapsed time for each.
• Status badges — see at a glance which are running, completed, or failed.
• Drill-in view — click any sub-agent to see its full chat transcript and task plan, rendered with the same components as the main chat.

When sub-agents are used

The agent decides to delegate based on:

• Independence — tasks that don’t depend on each other’s output.
• Complexity — the overall change is large enough to benefit from parallel execution.
• Specialization — different parts of the change require different expertise (copy vs. layout vs. logic).

You don’t need to ask for sub-agents explicitly. The agent uses them when they’ll make the work faster and more reliable.

What makes this different

Most AI coding tools run a single agent loop — one task at a time, sequentially. Sub-agents give Coframe Agent:

• Faster delivery — independent tasks complete in parallel.
• Better isolation — a failure in one sub-agent doesn’t block others.
• Full observability — you can watch each sub-agent’s work, not just the final merged result.
• Contextual focus — each sub-agent works with a narrower scope, producing more targeted changes.

Note

Sub-agents share the same workspace and file system as the parent agent. The agent coordinates writes so sub-agents don’t step on each other’s changes.