{
  "id": 167597,
  "title": "Claude Code Subagents: Setup, Config, and When to Use Them",
  "url": "https://urgent.news/2026/08/05/claude-code-subagents-setup-config-and-when-to-use-them",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-05T09:07:21.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/rosgluk/claude-code-subagents-setup-config-and-when-to-use-them-44m7"
  },
  "original_language": "en",
  "account": "Claude Code subagents offer a solution to manage noisy, parallelizable work by isolating it in an independent reasoning agent. This prevents the main conversation from becoming cluttered with extraneous information. A subagent is distinct from a Skill, which adds a capability to the main agent's context, and from the MCP server, which connects to external tools or data sources.\n\nTo effectively use subagents, it's crucial to understand their key characteristics: isolated context, restricted tool allowlist, and no cross-subagent visibility. When a task generates a lot of intermediate output, a subagent can help by delegating the work and only returning a summary. For tasks requiring tight back-and-forth refinement or sequential dependencies, a subagent is not the right solution. Instead, these should be handled sequentially.\n\nUsing subagents can significantly reduce the amount of context that the main session has to process. For instance, if you need to grep a large service for deprecated config keys, a subagent can generate a concise report with only the relevant findings. This approach can save a significant amount of context tokens and reduce costs. For example, using an Opus model for this task would cost around $0.20-0.25 for the input tokens alone, but by delegating to a subagent, the cost can be reduced to $0.04-0.05.\n\nCreating a custom subagent involves defining a YAML file with a name, description, tools, and optional skills. The description should clearly outline what the subagent is meant to accomplish. The tools section specifies the capabilities the subagent has access to, while the skills section lists any reusable procedures or playbooks it can utilize. Once defined, subagents can be shared across the team or used personally in each project.",
  "summary": "Most Claude Code sessions get slow and cluttered for the same reason: every exploratory grep, every log dump, and every \"let me check one more file\" stays in the main conversation forever. Subagents exist to fix exactly that problem. They are one of the agent primitives built into Claude Code for handling noisy, parallelizable work — a way to push the mess into an isolated window and bring back…",
  "key_points": [
    "Claude Code subagents isolate noisy work in independent reasoning agents",
    "Subagents have isolated context, tool allowlist, no cross-subagent visibility",
    "Subagents reduce main session context, save costs for tasks like grep"
  ],
  "editors_take": null,
  "illustration": "https://urgent.news/ill/167597.png",
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}