Deep Agentsを使用したマルチエージェントアプリケーションの構築

imageBy Sydney Runkle and Vivek Trivedy

Breaking down complex tasks across specialized agents is one of the most effective approaches to building capable AI systems.

Deep Agents makes this easy with two first-class primitives:

subagents: delegating to isolated agents

skills: progressively disclosing capabilities

In this post, we'll show you how to build multi-agent systems with Deep Agents.

Using Subagents: Specialized, Isolated Workers

Subagents tackle a fundamental problem in agent engineering: context bloat. This is when an agent’s context window becomes close to full as it works on a task.

Why is this important? There’s great work from Chroma on context rot showing that models struggle to complete tasks as their context window gets filled. Our friends at HumanLayer call this high context regime the “dumb zone”. Subagents isolate context from the main agent to help avoid quickly entering the dumb zone.

When your agent makes dozens of web searches or file reads, the context window fills with intermediate results. Subagents isolate work by running with their own context window. So if the subagent is doing a lot of exploratory work before coming with its final answer, the main agent still only gets the final result, not the 20 tool calls that produced it.

Here’s a look at the basic subagents architecture:

imageWhen to Use Subagents

Context Preservation: A task requiring multiple steps can clutter the main agent's context (ex: codebase exploration).

Specialization: Use domain specific instructions or tools. Subagents developed by distinct teams can specialize in different verticals.

Multi-Model: Subagents can use different models than the main agent. For example, choosing a smaller model for lower latency.

Parallelization: Subagents can run simultaneously and return their outputs to the main agent. This reduces latency.

Creating Subagents

Define subagents as dictionaries and pass them to create_deep_agent():

from deepagents import create_deep_agent

research_subagent = {

"name": "research-agent",

"description": "Used to research more in depth questions",

"system_prompt": "You are a great researcher",

"tools": [internet_search],

"model": "openai:gpt-4o", # Optional: override main agent model

}

agent = create_deep_agent(

model="claude-sonnet-4-5-20250929",

subagents=[research_subagent]

)

See the subagents documentation for configuration details.

The General-Purpose Subagent

Deep Agents include a built-in general-purpose subagent that mirrors your main agent's capabilities. It has the same system prompt, tools, and model. This is perfect for context isolation without specialized behavior.

Example: Instead of your main agent making 10 web searches and filling its context, it can delegate to the general-purpose subagent with task(name="general-purpose", task="Research quantum computing trends"). The subagent performs all searches internally and returns only a summary.

Best Practices for Subagents

Write clear descriptions. Your main agent uses descriptions to decide which subagent to call:

✅ Good: "Analyzes financial data and generates investment insights with confidence scores"

❌ Bad: "Does finance stuff"

Keep system prompts detailed. Include tool usage guidance and output format requirements:

research_subagent = {

"name": "research-agent",

"description": "Conducts in-depth research using web search and synthesizes findings",

"system_prompt": """You are a thorough researcher. Your job is to:

1. Break down the research question into searchable queries
2. Use internet_search to find relevant information
3. Synthesize findings into a comprehensive but concise summary
4. Cite sources when making claims

Output format:

• Summary (2-3 paragraphs)
• Key findings (bullet points)
• Sources (with URLs)

Keep your response under 500 words to maintain clean context.""",

"tools": [internet_search],

}

Minimize tool sets. Only give subagents the tools they need:

✅ Good: Focused tool set

email_agent = {

"name": "email-sender",

# Only email-related

"tools": [send_email, validate_email],

}

❌ Bad: Too many tools

email_agent = {

"name": "email-sender",

# Unfocused

"tools": [send_email, web_search, database_query, file_upload],

}

Using Skills: Progressive Disclosure of Capabilities

Skills provide a different pattern: progressive disclosure. Instead of giving your agent dozens of tools upfront, you define specialized capabilities in SKILL.md files. Your agent sees skill names and descriptions, then reads the full instructions only when needed.

imageSkill descriptions are pre-loaded into the context window. The skill body is only loaded when the agent decides the skill is needed based on the description and previous context.Caption: skill descriptions are pre-loaded into the context window. The skill body is only loaded when the agent decides the skill is needed based on the description and previous context.

Setting Up Skills

Skills use the agentskills.io spec. Here's the structure:

.deepagents/skills/

├── deploy/SKILL.md

└── review-pr/SKILL.md

Each SKILL.md file has YAML frontmatter with metadata and a main body:

name: deploy

description: Deploy to production

version: 1.0.0 # Optional

tags: [deployment, production] # Optional

Deploy to Production

When the user asks to deploy, follow these steps:

1. Run tests: npm test
2. Build the application: npm run build
3. Deploy to production: npm run deploy:prod
4. Verify deployment: Check the health endpoint

Always confirm with the user before deploying to production.

Adding Skills to Your Agent

Use the skills argument to create_deep_agent to load skills from the filesystem:

from deepagents import create_deep_agent

from deepagents.backends import FilesystemBackend

agent = create_deep_agent(

model="claude-sonnet-4-5-20250929",

backend=FilesystemBackend(root_dir="/"),

skills=[".deepagents/skills"],

)

The agent now sees your skills. When it needs detailed instructions, it reads the full SKILL.md file.

You can also use other backends (such as a StateBackend or StoreBackend), then invoke the agent with a files specification:

from deepagents.middleware.filesystem import FileData

default backend is a StateBackend

agent = create_deep_agent(

model="anthropic:claude-sonnet-4-20250514",

skills=["/skills/"],

)

skill_content = """

name: deploy

...

"""

Invoke the agent with the skill and virtual files

result = agent.invoke({

"messages": [HumanMessage(content="Research the latest Python releases")],

"files": {

"/skills/web-research/SKILL.md": FileData(

content=skill_content.split("\n"),

created_at="2024-01-01T00:00:00Z",

modified_at="2024-01-01T00:00:00Z",

),

},

})

Choosing the Right Pattern

Here’s a quick set of questions to guide you:

When you need to...

Use...

Both?

Delegate complex, multi-step work

Subagents for context isolation

Reuse procedures or instructions

Skills for progressive disclosure

Provide specialized tools for specific tasks

Subagents with focused tool sets

✅

Share capabilities across multiple agents

Skills (they’re just files)

✅

Work with large tool sets

Skills to avoid token bloat



Note, this doesn’t have to be an either-or decision.

Many systems use both. Skills define procedures; subagents execute complex multi-step work. Your subagents can use skills to effectively manage their context windows!

Next Steps

To learn more about multi-agent patterns in Deep Agents, check out our:

Subagents documentation - Detailed API reference and examples

Skills documentation - Detailed API reference and examples

Multi-agent patterns guide - General guidance on choosing patterns

The key insight: multi-agent patterns don't have to be complicated. With the right abstractions (middleware for plumbing, tool calling for invocation), they become simple building blocks you can compose into capable, sophisticated systems.

Start with subagents for context management, add skills for progressive disclosure, and build from there.