Wirewalk Instituteof AI
bld-320 Expert AvailableFlagship

AI Agents: Tool Use, MCP and Multi-Agent Orchestration

For engineers shipping LLM agents to production: design tools, build and secure MCP servers, choose an SDK, orchestrate multiple agents, and prove it works with evals.

Level
Expert
Study time
20 hours
Lessons
19 in 6 modules
Exam
25 questions

About this course

Agents fail in production for predictable reasons: vague tool interfaces, runaway context, unbounded permissions, multi-agent designs that multiply cost without improving results, and no way to tell whether a change helped. This course works through each of those failure points with code you can run.

You start with the agent loop itself and the wire formats for tool calling at Anthropic, OpenAI and Google, then design tool interfaces that models use correctly. You build and secure an MCP server against the current (2026-07-28) specification, including Streamable HTTP and OAuth 2.1. You compare the OpenAI Agents SDK, Claude Agent SDK, Google ADK and LangGraph on the same problems, and learn when a plain loop beats all of them. Later modules cover context and memory for long tasks, planning, orchestrator-worker and handoff patterns, parallel fan-out and when not to use it, computer and browser use, human approvals, prompt injection through tools, sandboxing, trajectory and outcome evaluation, and cost and latency control. Product facts are current as of October 2026.

What you will be able to do

  • Implement a tool-calling agent loop and translate tool definitions between the Anthropic, OpenAI and Gemini wire formats
  • Design tool interfaces with clear contracts, bounded outputs and actionable errors that reduce model misuse
  • Build an MCP server exposing tools, resources and prompts, and secure its Streamable HTTP endpoint with OAuth 2.1 audience-bound tokens
  • Select an agent SDK or a plain loop for a given workload using explicit criteria such as durability, approvals, tracing and model portability
  • Manage context and memory for long-running tasks using trimming, compaction, external memory files and explicit plans
  • Choose between single-agent, orchestrator-worker, handoff and parallel fan-out designs and justify the cost trade-off
  • Apply least privilege, human approval gates and sandboxing to defend agents against prompt injection delivered through tools
  • Build trajectory and outcome evaluations and use them to control cost and latency without losing quality

Syllabus

6 modules · 19 lessons · about 20 hours

Module 1: Agent Foundations and Tool Calling

What an agent actually is, how tool calls travel over the wire at each major vendor, and how to design tools models use correctly.
3 lessons
  1. What Makes an Agent: The Loop, the Harness and When Not to Build One
  2. Tool Calling Across Vendors: Anthropic, OpenAI and Gemini Wire Formats
  3. Designing Tool Interfaces Models Use Correctly

Module 2: The Model Context Protocol

MCP architecture under the stateless 2026-07-28 specification, building servers, and securing transports with OAuth 2.1.
3 lessons
  1. MCP Architecture: Hosts, Clients, Servers and the Stateless 2026-07-28 Revision
  2. Building an MCP Server: Tools, Resources and Prompts
  3. MCP Transports and Authorization: stdio, Streamable HTTP and OAuth 2.1

Module 3: Agent SDKs Compared

The OpenAI Agents SDK, Claude Agent SDK, Google ADK and LangGraph on the same problems, and how to choose.
3 lessons
  1. OpenAI Agents SDK and Claude Agent SDK in Practice
  2. Google ADK and LangGraph in Practice
  3. Choosing a Framework, or None: A Decision Matrix

Module 4: Memory, Context and Planning for Long Tasks

Keeping long-running agents coherent with context budgets, memory stores and explicit plans with stopping rules.
3 lessons
  1. Context Engineering for Long-Running Agents
  2. Memory Architectures: Scratchpads, Files and Long-Term Stores
  3. Planning, Replanning and Knowing When to Stop

Module 5: Multi-Agent Orchestration and Computer Use

Orchestrator-worker, handoffs and parallel fan-out, when a single agent is better, and agents that drive screens and browsers.
3 lessons
  1. Orchestrator-Worker, Handoffs and Agents-as-Tools
  2. Parallel Fan-Out, Aggregation and When Not to Go Multi-Agent
  3. Computer and Browser Use Agents

Module 6: Securing, Evaluating and Operating Agents

Approvals, least privilege, injection defenses, sandboxing, trajectory and outcome evals, and cost and latency control.
4 lessons
  1. Human Approvals and Least Privilege
  2. Prompt Injection via Tools, Sandboxing and the MCP Supply Chain
  3. Trajectory and Outcome Evals for Agents
  4. Cost and Latency Control for Agents

Exam and certificate

The exam unlocks once you have completed every lesson. It draws 25 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 60 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.

Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.

The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Anthropic, OpenAI, Google, LangChain, or Model Context Protocol. Product and company names are trademarks of their respective owners.

More in Building AI Applications

All 14 courses in this track
bld-101 Basic Waitlist

No-Code AI Assistants: Custom GPTs, Projects, Gems and Copilot Studio

A hands-on course on building no-code assistants with the builders already in your workplace tools: ChatGPT projects, custom GPTs and their successor plugins, Claude Projects and Skills, Gemini Gems, and Microsoft 365 Copilot Agent Builder and Copilot Studio. You leave with a tested, documented assistant and a plan to share it safely.

7 h · 17 lessons Join the waitlist
bld-201 Intermediate Available

Building with LLM APIs Across Vendors

A hands-on, code-first course on the three major LLM APIs side by side: messages, streaming, structured outputs, tool calling, vision, cost, caching, batch, retries, abstraction layers, cloud routes and testing. You finish by building and hardening a small multi-vendor service.

14 h · 18 lessons See price at checkout
bld-210 Intermediate Available

Retrieval-Augmented Generation Systems

A practitioner's course on retrieval-augmented generation: when to use it, how to ingest, chunk, embed and index documents, how to combine BM25 and vector search with reranking, and how to measure, debug and operate the result. Every module includes working Python sketches.

14 h · 16 lessons See price at checkout