AI Agent Code Library
Copy-ready starter code to create AI agents with the most popular SDKs and frameworks. Paste, add your API key, run.
export ANTHROPIC_API_KEY=… or OPENAI_API_KEY). Model names and SDK versions change often — check each project's docs for the latest.Raw tool-use loop — Claude API (Python)
The core pattern behind every agent: the model asks for a tool, your code runs it, the result goes back, repeat until done.
pip install anthropic · Docs ↗
import anthropic
client = anthropic.Anthropic() # reads ANTHROPIC_API_KEY
def get_weather(city: str) -> str:
return f"It's 68°F and sunny in {city}." # call a real API here
tools = [{
"name": "get_weather",
"description": "Get the current weather for a city.",
"input_schema": {
"type": "object",
"properties": {"city": {"type": "string"}},
"required": ["city"],
},
}]
messages = [{"role": "user", "content": "Should I bring a jacket in Monterey today?"}]
while True:
resp = client.messages.create(
model="claude-sonnet-5-5",
max_tokens=1024,
tools=tools,
messages=messages,
)
messages.append({"role": "assistant", "content": resp.content})
if resp.stop_reason != "tool_use":
print("".join(b.text for b in resp.content if b.type == "text"))
break
results = []
for block in resp.content:
if block.type == "tool_use":
output = get_weather(**block.input)
results.append({"type": "tool_result", "tool_use_id": block.id, "content": output})
messages.append({"role": "user", "content": results})
OpenAI Agents SDK (Python)
Define an agent with instructions and function tools; the Runner handles the loop.
pip install openai-agents · Docs ↗
from agents import Agent, Runner, function_tool
@function_tool
def get_weather(city: str) -> str:
"""Get the current weather for a city."""
return f"It's 68°F and sunny in {city}."
agent = Agent(
name="Weather helper",
instructions="Answer briefly and practically.",
tools=[get_weather],
)
result = Runner.run_sync(agent, "Should I bring a jacket in Monterey today?")
print(result.final_output)
Claude Agent SDK (Python)
The same harness that powers Claude Code — built-in file, shell and web tools, ready to use.
pip install claude-agent-sdk · Docs ↗
import asyncio
from claude_agent_sdk import query, ClaudeAgentOptions
async def main():
options = ClaudeAgentOptions(
system_prompt="You are a careful code reviewer.",
allowed_tools=["Read", "Grep", "Glob"],
)
async for message in query(prompt="Review the files in ./src for bugs.", options=options):
print(message)
asyncio.run(main())
Multi-agent team — CrewAI (Python)
Give each agent a role and goal, assign tasks, and let the crew work in sequence.
pip install crewai · Docs ↗
from crewai import Agent, Task, Crew
researcher = Agent(
role="Market Researcher",
goal="Find the top AI agent platforms for small businesses",
backstory="You research software markets and cite your sources.",
)
writer = Agent(
role="Content Writer",
goal="Turn research into a clear, friendly blog post",
backstory="You write plain-English guides for busy owners.",
)
research = Task(
description="List 5 AI agent platforms for small businesses with pros and cons.",
expected_output="A bullet list of 5 platforms with pros/cons.",
agent=researcher,
)
post = Task(
description="Write a 400-word blog post from the research.",
expected_output="A 400-word blog post in markdown.",
agent=writer,
)
crew = Crew(agents=[researcher, writer], tasks=[research, post])
print(crew.kickoff())
LangChain agent (Python)
A prebuilt tool-calling agent in a few lines; swap in any supported model.
pip install langchain langchain-anthropic · Docs ↗
from langchain.agents import create_agent
def get_weather(city: str) -> str:
"""Get the current weather for a city."""
return f"It's 68°F and sunny in {city}."
agent = create_agent(
model="anthropic:claude-sonnet-5-5",
tools=[get_weather],
system_prompt="You are a helpful assistant.",
)
result = agent.invoke({"messages": [{"role": "user", "content": "Weather in Monterey?"}]})
print(result["messages"][-1].content)
Build an MCP server (Python)
Expose your own tools to any MCP-compatible agent — Claude, ChatGPT, Cursor and more.
pip install "mcp[cli]" · Docs ↗
from mcp.server.fastmcp import FastMCP
mcp = FastMCP("listings")
@mcp.tool()
def search_listings(city: str, max_price: int) -> list[dict]:
"""Search property listings in a city under a max price."""
# Replace with a real database or API call
return [{"address": "123 Ocean Ave", "city": city, "price": max_price - 50000}]
if __name__ == "__main__":
mcp.run() # stdio transport by default
Tool-calling agent — Vercel AI SDK (TypeScript)
For web developers: a multi-step agent with typed tools in Node or Next.js.
npm i ai @ai-sdk/anthropic zod · Docs ↗
import { generateText, tool, stepCountIs } from "ai";
import { anthropic } from "@ai-sdk/anthropic";
import { z } from "zod";
const { text } = await generateText({
model: anthropic("claude-sonnet-5-5"),
stopWhen: stepCountIs(5),
tools: {
getWeather: tool({
description: "Get the current weather for a city",
inputSchema: z.object({ city: z.string() }),
execute: async ({ city }) => `It's 68°F and sunny in ${city}.`,
}),
},
prompt: "Should I bring a jacket in Monterey today?",
});
console.log(text);
How to Build an AI Agent
Define the job
One clear goal, the inputs it gets, and what “done” looks like.
Pick model + framework
Start with a vendor SDK; reach for LangGraph or CrewAI when you need complex flows or teams.
Write tools
Small functions with clear names, descriptions and typed inputs — or plug in MCP servers.
Add knowledge
Retrieval from a vector DB or a memory layer like Mem0 for long-term context.
Guardrails & evals
Limit permissions, require approval for risky actions, trace with Langfuse or LangSmith.
Deploy
Run in a sandbox (E2B), a serverless function, or a managed platform like Bedrock AgentCore.
Browse every tool mentioned here in Build Agents →