BRK570 Breakout Fri 20 Nov 09:00 PST
Developer's guide to Copilot Studio - Why and How
Matt Barbour
Summary
From the recording
The session presented Copilot Studio as a developer platform. Matt Barbour demonstrated the Ms. Agent CLI and a VS Code extension for cloning agents and workflows into Git, with dev, test and prod deployments, publishing and testing from the command line. He said the Copilot managed runtime hit GA this week. Git actions and integration for service principal-based deployments through the CLI have not yet shipped. He said he built the airport weather demo agent for about 30 bucks. Caution: the GA statement was hedged ('I believe'); service principal Git actions were only said to be coming 'relatively soon'.
Session description
Microsoft's catalogue description
You know how to build an agent in code. So what does Copilot Studio add? In this session, we’ll show how developers can bring familiar pro-code practices like GitHub repo-backed development and agentic editing to Copilot Studio's connectors, monitoring and no-code experience. Build what your org needs locally in Visual Studio or even a terminal, then hand off to business owners to maintain and manage in Copilot Studio. Come see how Copilot Studio can complement the way you already build!
Demo timestamps
Microsoft's player does not support links to a specific time, so open the recording and skip to the time shown.
- Skip to 00:14:34
Copilot Studio CLI in action
The CLI is already running: command help, listing environments and agents, and a local create of those capabilities, with the speaker asking the coding agent what it is doing.
- Skip to 00:24:28
Editing the same agent in Copilot Studio
The same agent (an NHL news desk sample in a preview environment) is edited in Copilot Studio in auto mode and saved, showing the CLI and studio staying in sync.
- Skip to 00:36:26
Deploying and testing the agent
The agent is deployed as live and test, with a discussion of interactive versus non-interactive testing.
- Skip to 00:43:20
UI preview making a text stream intelligible
A UI preview is used to make a text stream intelligible, with the speaker asking for the TAF for a weather site. Projects beyond the basic connector set are noted as possible.