Build & deploy agents in minutes

Quick Start

Create and deploy your first agent in under 5 minutes. This example creates a customer support agent with tool access.

$ npx agently init my-agent
# Creates a new agent project with config and boilerplate

from agently import Agent, deploy

# Define your agent
agent = Agent(
    name="support-bot",
    model="gpt-4o",
    instructions="You are a helpful customer support agent.",
    tools=["search_docs", "create_ticket"]
)

# Deploy to production in one line
result = deploy(agent, environment="production")
print(f"Agent live at: {result.url}")

Project Structure

An Agently project follows a simple, opinionated structure. The CLI generates this when you run agently init.

my-agent/
|-- agent.py           # Agent definition & tools
|-- agently.yaml       # Configuration & deploy settings
|-- tools/
|   |-- search.py      # Custom tool implementations
|   |-- tickets.py
|-- tests/
|   |-- test_agent.py  # Agent behavior tests
|-- requirements.txt   # Python dependencies
|-- .env               # Environment variables (local)

API Reference

Core API

The Agently SDK exposes three primary functions: create, deploy, and monitor. Each is designed to be simple to use while offering deep configurability when needed.

#Agent()— Create Agents

Instantiate an agent with a model, instructions, and optional tool integrations. Agents are stateless by default but can be configured with persistent memory.

Parameters

from agently import Agent

agent = Agent(
    name="my-agent",
    model="gpt-4o",
    instructions="Respond helpfully to user queries.",
    tools=["web_search", "calculator"],
    memory=True,
    max_tokens=4096
)

#deploy()— Deploy Agents

Deploy an agent to Agently's managed infrastructure. Handles containerization, networking, TLS, and auto-scaling automatically.

Parameters

from agently import deploy

result = deploy(
    agent,
    environment="production",
    scaling={ "min_instances": 1, "max_instances": 10 },
    region="us-east-1"
)

print(result.url)      # https://my-agent.agently.run
print(result.status)   # 'deployed'
print(result.version)  # 'v1'

#monitor— Monitor Agents

Access real-time logs, performance metrics, and execution traces for any deployed agent. Essential for debugging and optimizing agent behavior.

Available Methods

from agently import monitor

# Get real-time logs
logs = monitor.logs("support-bot", tail=100)

# Get performance metrics
metrics = monitor.metrics("support-bot")
print(metrics.avg_latency_ms)   # 245
print(metrics.requests_24h)     # 12847
print(metrics.error_rate)       # 0.002

# Get execution traces
traces = monitor.traces(
    "support-bot",
    limit=10,
    include_tool_calls=True
)

Code Examples

Full example: Customer support agent

End-to-end example showing how to create a customer support agent with custom tools, deploy it to production, and monitor its performance — all in under 30 lines.

from agently import Agent, deploy, monitor, Tool

# 1. Define custom tools
@Tool(name="search_knowledge_base")
def search_kb(query: str) -> str:
    """Search the company knowledge base for relevant articles."""
    # Your search logic here
    return perform_search(query)

@Tool(name="create_support_ticket")
def create_ticket(
    title: str,
    priority: str = "medium"
) -> dict:
    """Create a new support ticket in the system."""
    return { "id": "TKT-1234", "status": "created" }

# 2. Create the agent
agent = Agent(
    name="customer-support",
    model="gpt-4o",
    instructions="""
    You are a world-class customer support agent.
    Always search the knowledge base before answering.
    Create tickets for issues you cannot resolve directly.
    Be empathetic, concise, and solution-oriented.
    """
)

# 3. Deploy to production
result = deploy(
    agent,
    environment="production",
    scaling={ "min": 2, "max": 20 },
)
print(f"Live at: {result.url}")

# 4. Monitor in real-time
metrics = monitor.metrics("customer-support")
print(f"Latency: {metrics.avg_latency_ms}ms")
print(f"Uptime:  {metrics.uptime_pct}%")