How can we help you?

Microsoft Copilot Studio Guide

Microsoft Copilot Studio Guide

What it is

The Microsoft Copilot Studio integration connects your own Microsoft Copilot Studio agent — or an Azure AI Agent Service agent from Azure AI Foundry — to Velaro as a live data source behind your Velaro bot.

It is not an ingested knowledge base. Nothing is crawled or copied. When a visitor asks a question that needs your internal data, Velaro's bot forwards that question to your Copilot Studio agent in real time and uses the agent's reply to compose its answer. Your agent can reach LIMS systems, Power Automate flows, internal topics, and any custom logic you have already built in Copilot Studio.

Velaro stays in front: it keeps tone, brand, channel control, and translation. Your agent answers the data-bound questions.

Connecting an agent

Open Integrations → Copilot Studio / Azure AI Agent and use the Direct Line / REST tab to launch the setup wizard.

The wizard walks you through four steps — Connect (choose Copilot Studio or Azure AI Agent Service and enter your credentials), Verify (Velaro calls your agent live and shows the real result before anything goes live), Configure (name the agent and write its "when to use" description), and Finish (review and save). Nothing is eligible to receive live traffic until Verify passes and Finish is confirmed.

If you already have one or more agents connected, reopening the page first asks what you want to do — add another agent, edit an existing one, or start over. Starting over requires typing the agent's name to confirm, and the prior configuration is snapshotted to the Activity Log before anything is removed, so nothing is silently overwritten by reopening the wizard.

Two agent types are supported:

TypeWhat you provideWhere to find it
Microsoft Copilot StudioDirect Line secretAzure Portal → Bot Services → your bot → Channels → Direct Line → Show key
Azure AI Agent ServiceAzure AI Foundry endpoint, Agent ID, API keyAzure AI Foundry → your project → Keys and Endpoint; the Agent ID is on the agent

Once saved, secrets are never displayed or sent back to your browser again — editing an existing agent lets you change its name, description, and active/inactive state, but reconnecting with a new secret requires the Start Over path. If you also use SharePoint, OneDrive, or Teams features, connect Microsoft once on the Microsoft 365 page — that shared connection is reused across all Microsoft-powered Velaro features.

Data handling and PII

Every message forwarded to your connected agent — and every reply it sends back — passes through Velaro's standard message-scrubbing layer before it leaves or enters the conversation, the same protection used everywhere else in Velaro. Your agent only ever receives what it needs to answer the question; Velaro does not add its own storage or copy of your agent's responses beyond the normal conversation transcript. Every connection call is logged through Velaro's three-layer integration logging so any failure is visible in the Activity Log tab on this page, not silently dropped.

Works with your own AI model (BYOK)

The Copilot Studio integration is a tool your bot calls — it doesn't care which model is doing the reasoning. If your site uses Bring Your Own Key (BYOK) for its underlying AI model (for example, your own Anthropic key), Copilot Studio and Azure AI Agent connections keep working exactly the same way: the model picks when to call your agent, and your agent's own Microsoft/Azure token spend is entirely separate from your Velaro AI model billing either way.

The "When to use this" description

Each connected agent has a When should the bot use this? field. This is a plain-language description of the questions the agent can answer, for example: "for questions about lab sample status, LIMS records, or order tracking".

The bot's AI reads this description to decide when to query the agent instead of answering from its own knowledge index or another connected source. Be specific — clear descriptions produce accurate routing. You can edit it any time on the Copilot Studio page.

Adding it as a connected source

A connected agent appears as a selectable Connected agent (live) entry wherever a bot picks its knowledge.

In a bot's settings, the Knowledge Source picker lists your Azure Search indexes and your connected agents together. A bot with no traditional index can use a connected agent on its own. In a workflow AI node, the Knowledge Sources panel lets you tick indexes and connected agents per node.

Indexed sources are searched offline (chunked and embedded). Connected agents are queried live — the difference is shown clearly in the picker.

The workflow node

The Ask Microsoft Copilot Studio AI node drops the integration straight into a workflow. Add it from the workflow editor, pick an AI Config and a connected agent, and the bot answers general questions from Velaro's knowledge base while forwarding data-bound questions to your agent.

The Copilot Studio page also has a Build Starter Workflow button that creates a ready-made "Live Data Assistant" bot for you.

Live query, not ingested

Every answer from a connected agent is fetched in real time. There is no crawl, no embedding, and no copy of your data inside Velaro. This means answers are always current, and the agent enforces its own permissions. The bot forwards the question and uses the reply — it never shows raw agent output and never invents data.

Subscription

The Microsoft Copilot Studio integration is gated by the Copilot Studio subscription feature. If the integration is locked, contact your Velaro account manager to enable it on your plan.

Share: Email

Was this article helpful?