Skip to main content
Version: Next

Quick Start

Build and run your first AI agent with Agent Kernel in under 5 minutes!

Choose Your Framework​

Agent Kernel supports multiple frameworks. Pick the one you're most comfortable with:

OpenAI Agents Quick Start​

1. Install​

pip install agentkernel[openai]

2. Create Your Agent​

Create a file called my_agent.py:

from agents import Agent as OpenAIAgent
from agentkernel.cli import CLI
from agentkernel.openai import OpenAIModule

# Define your agent
general_agent = OpenAIAgent(
    name="general",
    handoff_description="Agent for general questions",
    instructions="You provide assistance with general queries. Give short and direct answers.",
)

math_agent = OpenAIAgent(
    name="math",
    handoff_description="Specialist agent for math questions",
    instructions="You provide help with math problems. Explain your reasoning.",
)

# Register agents with Agent Kernel
OpenAIModule([general_agent, math_agent])

if __name__ == "__main__":
    CLI.main()

3. Set API Key​

export OPENAI_API_KEY=your-api-key-here

4. Run Your Agent​

python my_agent.py

Testing Your Agent​

Once your agent is running, you'll see an interactive CLI:

(kernel) >> Using in-memory session store
AgentKernel CLI (type !help for commands or !quit to exit):
(kernel) >> No agent was requested. Defaulting to first agent in the list
(kernel) >> Selected agent: triage
(kernel) >> Starting new session: 07ec500e-f103-4b0e-8ecb-d794232f5992
(triage) >>
(triage) >> !list
Available agents:
  triage
  math
  general

(triage) >> !select math
(math) >> What is 2 + 2?

[math agent responds]
(math) >> 2 + 2 = 4. This is basic addition where we combine two quantities...

(math) >>

Understanding the Structure​

Every Agent Kernel application follows this pattern:

  1. Define agents using your preferred framework
  2. Wrap them in an Agent Kernel Module
  3. Run them using Agent Kernel's execution modes (CLI, API, AWS, etc.)

Next Steps​

Add Custom Tools​

Enhance your agent with custom tools:

from crewai import Agent, Tool

def search_database(query: str) -> str:
    # Your custom logic
    return f"Results for: {query}"

search_tool = Tool(
    name="search",
    description="Search the database",
    func=search_database
)

agent = Agent(
    role="researcher",
    goal="Find information",
    backstory="You are a research assistant",
    tools=[search_tool],
    verbose=False
)

Deploy as REST API​

Run your agent as a REST API server:

# Install API dependencies
pip install agentkernel[api]

Use the RESTAPI module run the agents

from agents import Agent as OpenAIAgent
from agentkernel.api import RESTAPI
from agentkernel.openai import OpenAIModule

# Define your agent
general_agent = OpenAIAgent(
    name="general",
    handoff_description="Agent for general questions",
    instructions="You provide assistance with general queries. Give short and direct answers.",
)

math_agent = OpenAIAgent(
    name="math",
    handoff_description="Specialist agent for math questions",
    instructions="You provide help with math problems. Explain your reasoning.",
)

# Register agents with Agent Kernel
OpenAIModule([general_agent, math_agent])

if __name__ == "__main__":
    RESTAPI.run()

Run the agents

# Run as API server
python my_agent.py

Configure the API port by editing config.yaml

api:
    port: 8000

Test it:

curl -X POST http://localhost:8000/run \
  -H "Content-Type: application/json" \
  -d '{"agent": "general", "message": "Hello!"}'

Deploy to AWS Lambda​

Package and deploy to AWS Lambda:

# Install AWS dependencies
pip install agentkernel[aws]

Use the Lambda module to run the agents

from agents import Agent
from agentkernel.aws import Lambda
from agentkernel.openai import OpenAIModule

math_agent = Agent(
    name="math",
    handoff_description="Specialist agent for math questions",
    instructions="You provide help with math problems. Explain your \
        reasoning at each step and include examples. \
        If prompted for anything else you refuse to answer.",
)

history_agent = Agent(
    name="history",
    handoff_description="Specialist agent for historical questions",
    instructions="You provide assistance with historical queries. \
        Explain important events and context clearly.",
)

triage_agent = Agent(
    name="triage",
    instructions="You determine which agent to use based on the user's question.",
    handoffs=[history_agent, math_agent],
)

OpenAIModule([triage_agent, math_agent, history_agent])

handler = Lambda.handler
# Deploy requires AWS credentials configured
# Terraform module imported and configured (see examples)
terraform init
terraform apply

Configure Memory​

Add Redis-backed memory for persistent sessions:

export AK_SESSION_STORAGE=redis
export AK_REDIS_URL=redis://localhost:6379

Or use in-memory storage (default):

export AK_SESSION_STORAGE=in_memory

Common Patterns​

Session Management​

# Sessions automatically track conversation history
# Each user/conversation gets a unique session ID
# Configure via environment variables:
# AK_SESSION_STORAGE=redis
# AK_REDIS_URL=redis://localhost:6379

Explore more examples in our Examples section:

Example snippets can be found here.

Troubleshooting​

Agent Not Found​

Make sure you're using the correct agent name when running:

# Agent names come from the framework-specific naming
# OpenAI: agent.name
# CrewAI: agent.role
# LangGraph: graph.name

API Key Errors​

Ensure your API key is set correctly:

# OpenAI (for OpenAI, CrewAI, LangGraph)
export OPENAI_API_KEY=sk-...

# Google (for ADK)
export GOOGLE_API_KEY=...

Import Errors​

Install the correct extras package:

pip install agentkernel[your-framework]

Learn More​

Need help? Check out our GitHub Issues or open a discussion!