BCS Blog

COBOL, z/OS performance measurement, and systems programming updates.

agent sdk

By Arch Brooks on July 14, 2026 in Programming

Back to Blog ← Back

Permalink

The OpenAI Agents SDK is an open-source framework for building AI agents that can reason, use tools, collaborate with other agents, and execute multi-step workflows. It sits on top of the Responses API and manages the orchestration so you don't have to write the control loop yourself.

Core concepts

Agent – An LLM configured with:
- Instructions (system prompt)
- Model (such as GPT-5)
- Tools (functions, APIs, web search, file access, etc.)
Tools – Python or JavaScript functions the agent can call.
Handoffs – One agent delegates work to another specialized agent.
Guardrails – Validate inputs and outputs, enforce policies, and improve safety.
Tracing – Visualize every reasoning step, tool call, and handoff to simplify debugging.

Typical architecture

                User
                  │
                  ▼
          ┌──────────────┐
          │ Main Agent   │
          └──────┬───────┘
                 │
      ┌──────────┼───────────┐
      ▼          ▼           ▼
 Search Agent  Coding     Database
               Agent       Agent
      │          │           │
      └──────────┴───────────┘
                 │
                 ▼
          Final Response

Simple Python example

from agents import Agent, Runner

agent = Agent(
    name="Assistant",
    instructions="You are a helpful programming assistant."
)

result = Runner.run_sync(
    agent,
    "Explain what RBAC is."
)

print(result.final_output)

Adding tools

from agents import Agent, function_tool

@function_tool
def get_weather(city: str):
    return f"It is sunny in {city}"

agent = Agent(
    name="Weather",
    instructions="Answer weather questions.",
    tools=[get_weather]
)

The agent decides when to invoke get_weather() automatically.

Multi-agent example

research = Agent(
    name="Research",
    instructions="Research technical topics."
)

coder = Agent(
    name="Coder",
    instructions="Write production-quality code."
)

manager = Agent(
    name="Manager",
    instructions="Delegate research and coding tasks.",
    handoffs=[research, coder]
)

JavaScript/TypeScript

import { Agent } from "@openai/agents";

const agent = new Agent({
  name: "Assistant",
  instructions: "You are a helpful assistant."
});

Newer capabilities

The latest versions of the SDK also support:

Native sandbox execution
File inspection and editing
Shell command execution
Long-running agent workflows
Checkpointing and resumable execution
Memory management
MCP integration
Skills support for reusable capabilities

When to use the Agents SDK

Use it when your application needs:

Multi-step reasoning
Tool calling
Multiple cooperating agents
Long-running workflows
Debugging with traces
Production-ready orchestration

If you only need a single prompt and response, using the Responses API directly is often simpler.

Documentation

Given your background as a C++ software engineer and your interest in local AI (Ollama, Whisper, Bootstrap/PHP applications), the Agents SDK is well suited for building:

Coding assistants
Multi-agent software engineering tools
Document analysis systems
Medical office assistants
Customer support agents
Autonomous workflow applications that combine local tools with cloud-based reasoning.

Comments

No comments yet.

Title *

Slug

Excerpt

Content *

HTML is supported.

SEO Meta Data

Meta Title

Meta Description