QuickStart Presets Guide

Rapid agent prototyping with zero infrastructure using pre-configured presets

QuickStart Presets

MCP Server LangGraph provides pre-configured presets for different deployment scenarios:

Preset	Setup Time	Infrastructure	Use Case
QuickStart	< 2 minutes	None	Learning, prototyping
Development	~15 minutes	Docker Compose	Local development
Production	1-2 hours	Kubernetes	Enterprise deployment

This guide covers the QuickStart preset for rapid agent development.

---

QuickStart Preset

The QuickStart preset provides:

• In-memory checkpointing (LangGraph MemorySaver)
• Free LLM defaults (Gemini Flash)
• No Docker required
• No authentication needed
• Simple agent creation
• FastAPI server ready

Quick Start

from mcp_server_langgraph.presets import QuickStart

# Create an agent in one line
agent = QuickStart.create("Research Assistant")

# Chat with the agent
result = agent.chat("What is LangGraph?")
print(result)

---

Creating Agents

Basic Agent

from mcp_server_langgraph.presets import QuickStart

# Create agent with default settings
agent = QuickStart.create("My Agent")

# Chat
response = agent.chat("Hello, how can you help me?")
print(response)

Agent with Tools

from mcp_server_langgraph.presets import QuickStart

# Create agent with tools
agent = QuickStart.create(
    name="Calculator Agent",
    tools=["calculator", "search"],
    llm="gemini-flash",
)

# Use the agent
result = agent.chat("What is the square root of 144?")

Agent with Custom Settings

from mcp_server_langgraph.presets import QuickStart

agent = QuickStart.create(
    name="Creative Writer",
    llm="claude-haiku",
    system_prompt="You are a creative writing assistant specialized in short stories.",
    temperature=0.9,  # Higher creativity
)

result = agent.chat("Write a short story about a robot learning to paint.")

---

Available LLMs

The QuickStart preset supports these free-tier friendly models:

Model	Provider	Best For
gemini-flash	Google	Fast, general purpose (default)
gemini-pro	Google	Complex reasoning
claude-haiku	Anthropic	Concise responses
gpt-5-mini	OpenAI	Balanced performance

# Use a specific LLM
agent = QuickStart.create(
    name="Assistant",
    llm="gemini-pro",  # Use Gemini Pro for complex tasks
)

---

Creating a FastAPI Application

Generate a complete REST API for your agent:

from mcp_server_langgraph.presets import QuickStart

# Create FastAPI app
app = QuickStart.create_app(
    name="Customer Support Bot",
    tools=["search", "knowledge_base"],
    llm="gemini-flash",
    port=8000,
)

# Run with: uvicorn app:app --reload

API Endpoints

The generated app includes:

Endpoint	Method	Description
/	GET	Health check with agent info
/chat	POST	Chat with the agent
/health	GET	Simple health check

Example Request

curl -X POST "http://localhost:8000/chat" \
  -H "Content-Type: application/json" \
  -d '{"query": "What can you help me with?", "thread_id": "user123"}'

Response:

{
  "query": "What can you help me with?",
  "response": "I'm Customer Support Bot, a helpful AI assistant...",
  "thread_id": "user123"
}

---

Conversation Threading

QuickStart supports conversation history with thread IDs:

agent = QuickStart.create("Assistant")

# First message in thread
response1 = agent.chat("My name is Alice", thread_id="user-123")

# Continuing the conversation
response2 = agent.chat("What's my name?", thread_id="user-123")
# Response will remember the context

---

Streaming Responses

For real-time output:

agent = QuickStart.create("Streaming Agent")

# Stream the response
for chunk in agent.stream_chat("Tell me a long story"):
    print(chunk, end="", flush=True)

---

Configuration Options

The QuickStartConfig model defines all options:

from mcp_server_langgraph.presets.quickstart import QuickStartConfig

config = QuickStartConfig(
    name="Custom Agent",
    tools=["search", "calculator"],
    llm="gemini-flash",
    system_prompt="You are a helpful research assistant.",
    temperature=0.7,
)

Option	Type	Default	Description
name	str	required	Agent name
tools	list[str]	[]	Tools to include
llm	str	gemini-flash	LLM model
system_prompt	str	None	Custom system prompt
temperature	float	0.7	LLM temperature (0.0-1.0)

---

Migrating to Production

When ready for production, migrate from QuickStart to the full deployment:

Step 1: Add Persistence

# QuickStart uses in-memory (data lost on restart)
# For production, use PostgreSQL checkpointing

# See: deployment/postgresql-checkpointing.mdx

Step 2: Add Authentication

# QuickStart has no authentication
# For production, enable Keycloak JWT auth

# See: guides/authentication.mdx

Step 3: Add Observability

# QuickStart has basic logging
# For production, add OpenTelemetry + LangSmith

# See: getting-started/langsmith-tracing.mdx

Step 4: Deploy to Kubernetes

# QuickStart runs locally
# For production, use Helm charts

helm install my-agent ./charts/mcp-server-langgraph

---

Limitations

The QuickStart preset is designed for learning and prototyping. It has these limitations:

Feature	QuickStart	Production
State persistence	In-memory (lost on restart)	PostgreSQL
Authentication	None	Keycloak JWT
Authorization	None	OpenFGA
Observability	Basic logging	OpenTelemetry + LangSmith
Scaling	Single instance	Kubernetes HPA
Secrets management	Environment variables	Infisical

---

Examples

Research Assistant

from mcp_server_langgraph.presets import QuickStart

agent = QuickStart.create(
    name="Research Assistant",
    tools=["search", "summarize"],
    llm="gemini-pro",
    system_prompt="You are a research assistant. Provide detailed, accurate information with sources.",
)

result = agent.chat("What are the latest developments in quantum computing?")

Code Helper

agent = QuickStart.create(
    name="Code Helper",
    llm="claude-haiku",
    system_prompt="You are a Python coding assistant. Provide concise, working code examples.",
    temperature=0.3,  # Lower temperature for more deterministic code
)

result = agent.chat("Write a function to calculate Fibonacci numbers")

Customer Support Bot

app = QuickStart.create_app(
    name="Support Bot",
    tools=["knowledge_base", "ticket_system"],
    llm="gemini-flash",
)

# Deploy with: uvicorn app:app --host 0.0.0.0 --port 8000

---

Related Documentation

• LangGraph Functional API
• Agent Architecture
• Deployment Overview