Enablement pack
A forkable repo, architecture write-up, runbook and walkthrough video.
Agentic Engineer. Build what agents run on.
Put an agentic UI on an internal system, with the scoped access, testing and deployment guardrails that make it safe to run. Assumes a software-developer base.
Self-paced · Free · No application
Technical pathOpen now
Four modules and a capstone
Sandbox your agents and set a quality bar, author and iterate agents, skills and memory, orchestrate and calibrate them, and govern what they can touch.
- 72–84 hours
- 30–90 days
By the end of this course, you will be able to:
- Run agents safely at scale
- Coding agents in containers, sessions in parallel, and a baseline every improvement is measured against.
- Orchestrate multi-agent pipelines
- Specialized subagents with scoped context and narrow tools, and their failures diagnosed and fixed.
- Govern what agents can touch
- Least-privilege policies for tools, skills and data, enforced in dev and CI/CD, every step right-sized for cost and speed.
Project-based, built for engineers doing the work
Learn by building something measurable every day. Move fast and ship in about a month—or fit it around a full-time job and take up to 90 days. Every module produces an artifact—baselines, a memory system, a routing-and-tools plan, guarded pipelines—that builds toward your capstone.
Module 01
Sandbox Your Agents and Set Your Quality Bar
- Lessons
- 3
- Exercises
- 4
- Hours total
- ~10
Stop running coding agents raw on your laptop. You will sandbox your coding harness in a container, run several agent sessions side by side without them stepping on each other, and define what "good" looks like with a product requirements document, a quality rubric, and baseline measurements.
What you'll learn
Sandbox coding agents in a container safe enough to scale
Run parallel agent sessions without collisions
Set a quality bar with a PRD, a rubric, and baseline measurements
Module 02
Author and Iterate Agents, Skills, and Memory
- Lessons
- 4
- Exercises
- 6
- Hours total
- ~14
An agent without memory starts every conversation from zero. You will build a layered memory system—versioned agents and skills, scoped context, and persistent storage—and use reflection on real sessions to keep it sharp instead of letting it drift.
What you'll learn
Author versioned agents and skills
Know what belongs in a skill, in context, and in long-term storage
Use reflection on real sessions to keep memory sharp
Module 03
Orchestrate, Equip, and Calibrate Your Agents
- Lessons
- 3
- Exercises
- 3
- Hours total
- ~15
This is where agents start coordinating themselves and accessing real data. You will direct subagent delegation inside one process—a planner spawning specialized subagents, each with scoped context and a deliberately narrow set of MCP-backed tools—and stand up MCP servers for persistent storage and vector retrieval. Then diagnose and fix failures like context bleed, routing misfires, and retrieval misses.
What you'll learn
Delegate work to specialized subagents with scoped context per role
Stand up MCP servers for persistent storage and vector retrieval
Diagnose orchestration failures and keep a calibration log of what you tuned and why
Module 04
Govern Your Agents and Right-Size Each Step
- Lessons
- 3
- Exercises
- 4
- Hours total
- ~9
Agents are powerful enough that "trust by default" is no longer safe. You will write least-privilege policies that pin down which tools, skills, and data each agent can touch—enforced in your dev environment and CI/CD—then audit a workflow to find steps where a few lines of deterministic code beat an agentic call.
What you'll learn
Write least-privilege policies for tools, skills, and data
Enforce policies in dev and CI/CD with an audit trail for every agent action
Match each step to the cheapest, simplest tool that does the job
Capstone project
Ship Your Agentic Engineering Paved Road
This is the artifact you hand to your boss or a hiring manager: a forkable enablement pack that proves you can build a containerized, governed, multi-agent engineering pipeline. Measure impact against your Module 1 baseline, capture your architecture decisions, record a short walkthrough, and write a one-pager a non-engineer can understand—a portfolio piece an interviewer can evaluate in five minutes and probe in thirty.
A governed pipeline
Containerized agents, routing and policy controls.
Proof of impact
Holdout evaluation and an impact report against your baseline.
Make the case
Architecture decisions and a stakeholder one-pager.
Before you enroll
Think you can skip AI 101?
Already building with AI—and comfortable in Git, containers, and CI/CD? Take the placement quiz to see whether you're ready to start here.
Take the placement quiz→
20 questions · ~6 minutes
Pick a pace that fits your week
Agentic Engineer is self-paced. Here is what each plan asks of a week.
30
days
Heads-down. Best if you can protect evenings and a weekend block.
Hours per week
- Agentic Engineer
- 17–20 hrs
60
days
The pace most learners pick. Fits either side of a full-time job.
Hours per week
- Agentic Engineer
- 8–10 hrs
90
days
The lightest weekly load. Best if your weeks are unpredictable.
Hours per week
- Agentic Engineer
- 6–7 hrs
From learning to earning
Some learners use this where they already work. For anyone who wants the next job, the path is three steps.
Complete a specialized track and submit your capstone.
You enter LaunchCode's candidate pool.
We match you with employers rebuilding their workflows.
Placement depends on employer demand and on the work you produce. Cost will never be the reason you do not get there.
500+ hiring partners
Hire Human is an initiative of LaunchCode, the nonprofit that has spent more than a decade placing people into technical roles at companies that do not hire on a hunch. The same network hires from these tracks, and we are placing graduates weekly.
Why it costs you nothing
LaunchCode is a 501(c)(3) nonprofit, built to get people into technology careers without asking them to pay for the privilege. Employers pay us when they hire, and that is what funds the training. You never pay us to learn, and you never pay us to be matched with an employer.
No tuition
Not now, not at the end.
No deposit
Nothing to put down to hold a seat.
No income share
Nothing taken from your pay, ever.
Sponsored by
Start today. Finish with something you built.
Self-paced, open now, and free for good.
Questions before you start? Email enrollment.hirehuman@launchcode.org.
What happens next
Today
A confirmation email with your course login.
Your first week
You finish your first lesson, its quiz, and your first hands-on lab.
When you finish your course
A capstone you keep, and a completion certificate.
What people ask before they enroll
Do I need a degree or a coding background?
No degree for any of them. AI 101 and Agentic Specialist assume no coding at all. Agentic Engineer expects real developer experience. You also need to be 18 or older.
Agentic Specialist curriculum →Agentic Engineer curriculum →
Can I do these courses if I am not looking for a new job?
Yes. Plenty of people take these to automate work at the job they already have. You decide whether you want placement, and nothing happens unless you ask for it.
Can I skip AI 101?
Yes. The placement quiz tells you whether AI 101 would mostly be review for you. It is a recommendation, not a gate, so you can enroll in a specialized track either way. It takes about five minutes.
How long does it take?
AI 101 is 51 to 57 hours, Agentic Specialist 64 to 70, and Agentic Engineer 72 to 84. You choose whether that lands across 30 days or 60, and Agentic Engineer adds a 90-day plan. Every course is self-paced, so you can stretch the calendar.
Is it really free?
Yes, and there is no upsell at the end. LaunchCode is a 501(c)(3) nonprofit, so we are not selling you a course. Employers pay us when they hire the people we train, and that pays for the courses.
Will I get a job?
If you want placement, finishing a specialized track puts you in LaunchCode's candidate pool, where we match you with employers at no cost. We place graduates weekly. The hiring decision is always the employer's.
What do graduates earn?
We do not have that figure yet. We will publish one for these tracks once enough placements are behind it.
Is there a certificate?
You get a completion certificate, but it is not an accredited degree. The capstone is what proves you can do the work.
Here to hire? See how teams hire from these tracks.