Copilot Studio Foundations — GitHub Copilot Harness

Build an agent that plans its own steps, reads documents and produces finished work — then prove, publish and cost it. No coding.

By the end of twelve hours, every participant has built, tested, evaluated, published and cost-modelled an autonomous agent that completes a multi-step business task from end to end. It reads an unseen supplier pack, reasons across it against internal policy, looks up what it does not know, writes its review as a document, delegates a compliance check to a second agent, and escalates what it cannot resolve. Participants never draw a flowchart: they write a brief, set boundaries, and let the agent plan its own route.

Innovation and automation teams facing processes too variable to capture as a fixed flowchart.

Business analysts and process owners who want to express a process as intent and constraints, not a decision tree.

Developers new to Copilot Studio, before multi-agent design and evaluation at scale.

Technical managers and architects who need to know what autonomy costs, what it risks, and where it is the wrong tool.

Agents that must behave identically every time, such as FAQs, fixed intake forms or compliance scripts. Our Foundations — Standard Harness course covers those.

Multi-agent topology or evaluation at scale. Our Practitioner — GitHub Copilot Harness course covers those.

ALM, security, governance or estate-wide licensing. Our Copilot Studio Enterprise course covers those.

Organisations that cannot allocate Copilot Credits for training: building on this harness consumes them from the first agent.

Duration: 2 days or 6 sessions × 2 hours
Level: Foundation · GitHub Copilot Harness
Delivery: Remote or on-site
Group size: Up to 12
Price: Priced per team — ask us

Get dates and a quote

We reply within one working day. No sales sequence.

What your team will be able to do

Outcomes, not topics covered.

By the end, each of these is something they have done rather than watched.

Explain what an autonomous agent is, and where fixed, rule-based automation is still the better tool.

Justify choosing the GitHub Copilot harness over the standard and Copilot Chat harnesses, including when not to.

Configure all seven Build tab components: instructions, knowledge, tools, skills, model, connected agents and memory.

Write instructions that set role, scope, question protocol, guardrails and output expectations.

Ground an agent in approved knowledge sources and verify its answers through citations.

Give an agent the ability to act, through tools, workflows and MCP servers.

Package reusable behaviour as skills, delegate specialist work to a connected agent, and judge when one agent should become several.

Build and run an evaluation set, and explain why one good test run proves little on this harness.

Publish, monitor and cost-model an agent, and name the design choices that drive credit consumption.

What's covered

10 modules, a little under half of it building.

Labs run on your own tenant, so whatever gets built stays with you. On this harness, building consumes Copilot Credits from the first agent, so a credit limit is set before anything is created. The few capabilities still in preview are kept to one module, so the course holds if they change. We publish what each day produces rather than a minute-by-minute timetable.

Day 1

Reason and build

Leaves with: An agent that reads your documents, plans its own steps and produces finished work 5 modules 2 hands-on

01

Why autonomous agents

What changes when an agent plans its own steps, where autonomy earns its cost, and where it is the wrong tool. Closes with a live run of the finished agent.
02

The harness decision, and the cost that follows

The three harnesses compared, and why on this one credits are consumed from the moment building starts.
03

The Build tab: seven components

Instructions, knowledge, tools, skills, model, connected agents and memory, and which one to change when behaviour is wrong.
04

Lab 1: Instructions and knowledge

A credit limit first, then instructions with guardrails, three knowledge sources and verified citations.
Hands-on
05

Lab 2: Files and multi-step reasoning

The agent analyses a supplier pack it has never seen, writes its review as a document, and recovers from a degraded input.
Hands-on

Day 2

Extend, prove and ship

Leaves with: An agent that acts in real systems, delegates, is proven reliable, and pays for itself 5 modules 2 hands-on

06

Tools, workflows and MCP servers

How an agent reaches outside itself, how to choose between the three, and how to design for failure.
07

Lab 3: Give the agent hands

An external lookup, a workflow that records the result, and instructions that govern when each is used.
Hands-on
08

Skills, connected agents and memory

Reusable skills, delegating to a specialist agent, and deciding what memory may keep. The preview capabilities are gathered here.
09

Lab 4: Delegate and remember

Extract a skill, build and connect a compliance agent, set memory boundaries, and run the whole task end to end.
Hands-on
10

Evaluate, publish, monitor and cost

An evaluation set that runs, publication and monitoring, a credit model for the sponsor, and the capstone.

Upcoming batches

No public batch scheduled yet.

Teams book this course on their own dates. Joining as an individual? Tell us and we'll let you know when the next batch opens.

Before the first day

What we need from you.

Prerequisites

No coding, no AI background, no prior Copilot Studio exposure, and no need to take our Standard Harness course first.
Copilot Studio access with the GitHub Copilot harness enabled in the tenant.
Permission to create an environment and an agent in that tenant.
A Copilot Credit allocation, and permission to set a credit limit: building consumes credits on this harness.
A SharePoint site the participant can read.
Every participant completes a five-minute access check 48 hours before day one.

We send a checklist a fortnight ahead and will talk to your platform team directly if access needs arranging.

Next step

Tell us about your team.

Roughly how many people, what they build today, and when you'd want to run it. We come back with dates and a quote.

Check availability