deep-agent-creation

Deep Agent Creation

Create and configure Deep Agents using create_deep_agent() with tools, system prompts, and model selection.

When to Use

• When creating a new Deep Agent from scratch
• When adding custom tools to an agent
• When configuring model and system prompt
• When setting up the main agent entry point

Available Operations

1. Install Deep Agents SDK
2. Define custom tools
3. Create agent with create_deep_agent()
4. Run agent with invoke() or stream()
5. Configure model provider

Multi-Step Workflow

Step 1: Install

pip install deepagents
# or
uv add deepagents

# For search capability
pip install tavily-python

Step 2: Set API Keys

export ANTHROPIC_API_KEY=your-key
# or
export OPENAI_API_KEY=your-key

Step 3: Define Custom Tools

# agent/tools.py
from langchain.tools import tool

@tool
def search_codebase(query: str, file_pattern: str = "**/*.py") -> str:
    """Search the codebase for files matching a pattern and containing a query."""
    import glob
    results = []
    for filepath in glob.glob(file_pattern, recursive=True):
        with open(filepath) as f:
            content = f.read()
            if query.lower() in content.lower():
                results.append(f"Found in {filepath}")
    return "\n".join(results) if results else "No matches found"

@tool
def run_tests(test_path: str = "tests/") -> str:
    """Run pytest on the specified path and return results."""
    import subprocess
    result = subprocess.run(
        ["pytest", test_path, "-v", "--tb=short"],
        capture_output=True, text=True, timeout=120
    )
    return result.stdout + result.stderr

@tool
def lint_code(path: str = "src/") -> str:
    """Run ruff linter on the specified path."""
    import subprocess
    result = subprocess.run(
        ["ruff", "check", path],
        capture_output=True, text=True
    )
    return result.stdout or "No lint issues found"

Step 4: Create the Agent

# agent/main.py
from deepagents import create_deep_agent
from agent.tools import search_codebase, run_tests, lint_code

agent = create_deep_agent(
    model="anthropic:claude-sonnet-4-6",
    tools=[search_codebase, run_tests, lint_code],
    system_prompt="""You are a coding assistant that helps developers write,
    test, and review code. You follow clean architecture principles and
    always run tests after making changes."""
)

# Run
config = {"configurable": {"thread_id": "session-1"}}
result = agent.invoke(
    {"messages": [{"role": "user", "content": "Review the auth module"}]},
    config=config
)
print(result["messages"][-1].content)