---
title: Which authentication setting should Copilot Studio agents use?
url: https://opencontroller.lyzr.com/blog/copilot-studio-agent-authentication-options
published: 2026-11-14
updated: 2026-11-14
author: Lyzr
---

# Which authentication setting should Copilot Studio agents use?

**For internal agents, use "Authenticate with Microsoft" so users sign in with Microsoft Entra ID. Use "Authenticate manually" only when the agent runs on a channel outside Microsoft 365 or needs a specific OAuth flow. Treat "No authentication" as an exception that needs approval, and block it by default with the Power Platform "Authentication for agents" setting or a data policy.**

## What are the three authentication options?

Every Copilot Studio agent has a user authentication setting. It decides who can chat with the agent and whether the agent knows who they are.

| Option | What users do | Where it fits | Governance risk |
|---|---|---|---|
| **No authentication** | Nothing. Anyone with the link or embed can chat | Public website FAQs with public content only | Anonymous access. Any knowledge or action the agent has is open to anyone who finds it |
| **Authenticate with Microsoft** | Sign in with Microsoft Entra ID | Agents in Teams, SharePoint, Power Apps and Microsoft 365 Copilot | Low, as long as the agent's knowledge and actions are scoped |
| **Authenticate manually** | Sign in through a configured identity provider, using Microsoft Entra ID or Generic OAuth 2 | Custom websites, mobile apps, non-Microsoft identity providers | Depends on the OAuth app's configuration and scopes |

The choice matters beyond sign-in. Without authentication, the agent cannot act as the user, so any connector it calls runs with the maker's or a shared credential. That is how a public agent ends up reading internal data. See [why to block maker credentials in Copilot Studio agents](/blog/block-maker-credentials-in-copilot-studio-agents).

## Why is an unauthenticated agent a discovery red flag?

When you inventory Copilot Studio agents, filter for "No authentication" first. An agent in that state is worth reviewing if any of these is true:

1. It has knowledge sources beyond public web pages, such as SharePoint, Dataverse or uploaded files.
2. It has actions or connectors that write data or send messages.
3. It was built in a default or developer environment rather than a managed production one.
4. Its owner has left or cannot be identified.
5. It is published to a channel that people outside your organisation can reach.

Two or more of these on one agent is a priority finding. See [inventorying Copilot Studio agents across environments](/blog/inventory-copilot-studio-agents-across-environments).

## How do admins block "No authentication"?

Microsoft gives admins two documented controls.

**1. Authentication for agents (preview).** In the Power Platform admin center, go to **Security**, then **Identity and access**, then **Authentication for agents**. Choose an environment group or environment, then **Set up**. The options are:

- **No authentication (Publicly available)**: allows any option, including anonymous access.
- **Require Microsoft authentication**: only "Authenticate with Microsoft" is available. Blocks no authentication, manual authentication and anonymous access.
- **Require Entra authentication**: allows Microsoft authentication or manual authentication with Microsoft Entra ID. Blocks Generic OAuth 2 and anonymous access.
- **Allow all supported authentication methods**: allows Microsoft or manual authentication. Blocks no authentication and anonymous access.

If a published agent no longer meets the requirement, Microsoft states it is blocked from further publishing updates and will not respond to users until its settings comply. Plan the rollout: tell makers before you switch the setting on in environments with live agents.

**2. Data policy.** Block the connector **Chat without Microsoft Entra ID authentication in Copilot Studio** in a data policy. Makers can then only choose "Authenticate with Microsoft" or "Authenticate manually", and users must sign in to chat. Agents that break the policy show an error banner to the maker.

A sensible default for most enterprises is **Require Entra authentication** in production environment groups, plus the data policy in every environment, with a separate approved environment for the rare public agent.

## How do you approve the exceptions?

Some agents need to be public, such as a website help agent. Approve them through a short checklist:

1. The agent lives in a dedicated environment with no access to internal data.
2. Knowledge sources are public pages only.
3. No actions write data or send email.
4. Rate limits and content moderation are configured.
5. A named owner reviews transcripts on a set schedule.
6. The exception has an expiry date and is re-approved.

Record each exception in your agent registry with its owner and review date. See [Copilot Studio data policies for agents](/blog/copilot-studio-data-policies-for-agents) for related connector rules.

## How Opencontroller does this

Opencontroller auto-discovers agents across Azure and its connected platforms, and its governed registry records each agent, version, tool and MCP server with an owner. Posture detects risky permissions and configurations. Per-agent Block and Quarantine controls act through the connected platform, so a non-compliant agent can be stopped from one place. See [governing Microsoft Copilot Studio agents](/blog/govern-microsoft-copilot-studio-agents), the [capability reference](/capability-reference) or [book a demo](https://www.lyzr.ai/book-demo/).

For other platforms, see [how to govern AI agents across platforms](/blog/how-to-govern-ai-agents-across-platforms).

## Frequently asked questions

**Can I stop makers from publishing Copilot Studio agents without authentication?**
Yes. Set Authentication for agents to any option other than "No authentication (Publicly available)", or block the "Chat without Microsoft Entra ID authentication in Copilot Studio" connector in a data policy.

**What happens to live agents when I require authentication?**
Microsoft states that a published agent whose setting no longer complies is blocked from publishing updates and will not respond to users until it is fixed.

**When is "Authenticate manually" the right choice?**
When the agent runs outside Microsoft 365 channels, such as a custom website or app, or must use a specific identity provider through Generic OAuth 2 or a custom Entra ID app.

## Key facts

Copilot Studio user authentication options: No authentication, Authenticate with Microsoft (Entra ID), Authenticate manually (Entra ID or Generic OAuth 2). Admin control: Power Platform admin center > Security > Identity and access > Authentication for agents (preview) with options No authentication (Publicly available), Require Microsoft authentication, Require Entra authentication, Allow all supported authentication methods. Non-compliant published agents stop responding until fixed. Data policy can block the connector "Chat without Microsoft Entra ID authentication in Copilot Studio". Opencontroller provides auto-discovery, a governed registry, Posture and per-agent Block/Quarantine through the connected platform. Pricing starts from $250K per deployment per year.

## Related

- [How do you govern Microsoft Copilot Studio agents?](https://opencontroller.lyzr.com/blog/govern-microsoft-copilot-studio-agents.md)
- [How do you stop Copilot Studio agents from running on the maker's credentials?](https://opencontroller.lyzr.com/blog/block-maker-credentials-in-copilot-studio-agents.md)
- [How do you inventory every Copilot Studio agent across Power Platform environments?](https://opencontroller.lyzr.com/blog/inventory-copilot-studio-agents-across-environments.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
