Hackages Logo

2-Day Hands-On Program · Remote or On-site

How to Build Reliable AI Agents

Testimonial 1
Testimonial 2
Testimonial 3
Testimonial 4
5.0
Trusted by 5000+ engineers

OVERVIEW

From impressive demos to agents you can ship and maintain

AI agents are easy to demo, but hard to make reliable in production. This two-day program uses the Vercel AI SDK as the backbone: Day 1 is fundamentals — generation and streaming, structured outputs, tool calling, and safe agent-shaped loops with real error handling. Day 2 goes deeper with MCP, memory and workflow patterns, middleware, then a guided build to deploy a real agentic feature (for example on Next.js), plus testing and telemetry so your team can iterate after the workshop.

WORKSHOP CURRICULUM

Two days — AI SDK fundamentals, then build & deploy

Content follows the AI SDK Core docs (generation, streaming, structured data, tool calling, MCP, middleware, testing, telemetry) and agent patterns your team can apply in production — not toy scripts.

WorkshopHands-on practice & coding
LectureConcepts & theory
Lecture
Agents vs. plain LLM apps: mental model, providers, and where Vercel AI SDK fits
Workshop
Generating text & streaming — latency, stops, and shaping responses
Workshop
Structured outputs & validation — schemas, parse failures, and safe defaults
Workshop
Tool calling — defining tools, schemas, handling bad or empty arguments
Workshop
Agent-shaped loops — model → tool → model, max steps, when to stop
Lecture
Errors & reliability basics — timeouts, retries, logging (intro to telemetry)

OUTCOMES

By the end of the program

For software developers, technical leads, and builders moving from experiments to dependable agentic systems — especially teams standardizing on the Vercel AI SDK and modern web stacks.

Solid grounding in AI SDK Core: generation, streaming, structured data, and tool calling
Hands-on agent loops with explicit limits, validation, and error handling
Exposure to MCP and workflow patterns you can extend to your own systems
A deployable slice: streaming agent with tools, ready to iterate after the workshop
Practices for testing and light telemetry so behavior doesn’t regress silently

PREREQUISITES

What you should already be comfortable with

Comfortable with JavaScript or TypeScript and everyday Git workflows
Familiarity with HTTP APIs and deploying a small web or API app (helpful for Day 2)
Prior LLM experience optional — we build from the SDK up

MEET YOUR INSTRUCTORS

Davy Engone

Davy Engone

Founder & Software Mentor @ Hackages

Davy helps teams ship software with confidence — including AI-assisted development and agentic systems. This program emphasizes reliability, evaluation, and shipping over slide-ware.

HOW WE RUN THE TWO DAYS

Sample structure — exact times can shift with your timezone and breaks

Day 1 · Morning

Providers & models, text & streaming, structured outputs — exercises with AI SDK Core

Day 1 · Afternoon

Tool calling and multi-step loops; failures, retries, and a reliability checklist

Day 2 · Morning

MCP, memory & prompts, workflows / subagents, middleware — toward a real feature

Day 2 · Afternoon

Build & deploy an agentic slice; testing, telemetry, evals, and Q&A

FOCUS AREAS

What we emphasize technically

SDK

Vercel AI SDK Core

generateText, streaming, structured outputs, tool calling, and provider setup — the spine of the exercises.

MCP

Tools & MCP

First-party tools plus Model Context Protocol for real integrations with clear boundaries.

Ship

Next.js & deploy

Typical stack for Day 2: App Router patterns, streaming routes, and shipping to a host like Vercel.

PRICING

Train your team today

The cost to train your team is a fraction of the cost to replace them.

€8,000/team

* excl VAT

What's included:

2 full days of hands-on learning (teams up to 15 people)
Delivered remotely or on-site
Curriculum aligned with Vercel AI SDK Core & agents patterns
Day 2 build track toward something you can deploy and extend

* Team pricing starts at €8,000 for up to 15 participants · may vary by format and location

FAQ'S

if your question wasn't answered below

Chances are at this stage you have a couple questions. Here are a few questions that are often asked.

Day 1 locks in fundamentals (tools, loops, structured outputs, failure modes). Day 2 goes deep with MCP, workflows, and a concrete build-and-deploy path so the team leaves with more than a demo.