Edit

Set up a managed connector

Create an authenticated connection to an external service (like Jira, Slack, or Outlook) with per-tool approval controls and parameter locking. Your agent gains access to only the operations you enable with the governance settings you configure.

Prerequisites

  • An agent in the Azure SRE Agent portal
  • SRE Agent Author or Administrator role on the agent (see User roles)
  • An account on the service you want to connect (for example, a Jira account, Slack workspace membership, or Microsoft 365 account)

Step 1: Open the managed connector wizard

  1. Open your agent in the Azure SRE Agent portal.
  2. In the left sidebar, expand Builder.
  3. Select Connectors.
  4. Select Add connector.

The connector picker opens, showing available connectors organized by tab: Telemetry, Notification, Code Repository, MCP, Incidents, Deployment, and Other.

Managed connectors show a Preview badge. Use the search box to find a connector by name across all tabs.

Step 2: Select a managed connector and authenticate

Select the connector card for the service you want to connect, then select Next.

The authentication step appears. What you see depends on the connector:

OAuth connectors (most services)

For Slack, Box, Confluence, OneDrive, SharePoint, and similar services:

  1. Select Sign in (or Sign in to [Service]).
  2. An OAuth popup opens. Sign in with your account for that service.
  3. When the popup closes, the connection card shows a green checkmark and your authenticated name.

Multi-auth connectors (Jira, GitHub, Azure DevOps)

Some connectors offer multiple authentication methods:

  1. Open the Authentication method dropdown.
  2. Choose your preferred method (for example, Oauth or API Token for Jira).
  3. Complete the sign-in or enter your credentials.

API key connectors (GitLab, Notion)

  1. Enter your API key in the text field.
  2. Select Connect.

Credential connectors (SQL Server, PostgreSQL)

  1. Enter the server address, database name, username, and password.
  2. Select Connect.

After successful authentication, select Next to continue.

Tip

If the OAuth popup doesn't appear, check your browser's popup blocker. Allow popups from sre.azure.com.

Step 3: Select operations

The operations step shows every available tool for the connector as a grid of clickable cards.

  1. Select individual operation cards to select or deselect them.
  2. Use Select all to enable every operation at once.
  3. Use the search box to filter operations by name.

The count badge (for example, 12 / 47) shows how many operations you selected out of the total available.

Note

Select only the operations you need. Fewer tools means better governance and more accurate agent behavior.

Select Next when you're done selecting operations.

Step 4: Configure governance

The governance step shows a three-column table for every selected operation:

Column What it controls
Tool Name Which tools are enabled (toggle on/off)
Parameter Policy Whether to lock parameter values
Permission Whether the agent needs approval to run this tool

Lock parameter values (Parameter Policy)

  1. Toggle the Parameter Policy switch ON for a tool.
  2. The tool expands to show all its parameters.
  3. Type a value in any parameter field to lock it. The agent uses this exact value every time.
  4. Leave a field empty for the agent to fill it at runtime based on context.

For more on how parameter locking works, see Managed connectors.

Set approval requirements (Permission)

For each tool, toggle the Permission between:

  • Allow: Agent executes freely (default). Use for read operations.
  • Ask: Agent pauses and requests your approval before each execution. Use for write operations that create, modify, or delete resources.

Warning

In Autonomous mode (scheduled tasks, HTTP triggers), every tool set to Ask executes without human approval. Only enable Autonomous mode for workflows where you trust all enabled operations. See Run modes.

Select Next when governance is configured.

Step 5: Review and create

The review step shows a summary:

  • MCP Server name (auto-generated)
  • Connector name and icon
  • Operation count and parameter count
  1. Review the summary.
  2. Select Create.

The connector is created immediately. It appears in your connectors list and your agent can start using the enabled tools right away.

Verify the connector works

After creating the connector:

  1. Check the connectors list. Your new connector should show a Connected status.
  2. Open a chat with your agent and ask it to use one of the enabled operations.

For example, if you connected Jira with "Get Issue" enabled:

"Get the details of Jira issue OPS-1234"

Edit or delete a connector

Edit

Select an existing managed connector to reopen the wizard. You can change which operations are enabled, update parameter values, and modify approval settings.

Delete

To remove a connector, select it from the connectors list and select Delete.

Set up a managed connector with the REST API

You can also create and manage managed connectors programmatically using the REST API.

1. Get a token

TOKEN=$(az account get-access-token --resource https://azuresre.dev --query accessToken -o tsv)
ENDPOINT="https://{agentEndpoint}"

2. List available connectors

curl -s -H "Authorization: Bearer $TOKEN" \
  "$ENDPOINT/api/v2/connectorV2/managedApis"

3. Get operations for a connector

curl -s -H "Authorization: Bearer $TOKEN" \
  "$ENDPOINT/api/v2/connectorV2/connectors/jira/operations"

4. Create a connection (OAuth connectors require portal sign-in)

curl -X PUT -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  "$ENDPOINT/api/v2/connectorV2/connections/jira" \
  -d '{ "properties": {} }'

5. Create the MCP server config with governance

curl -X PUT -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  "$ENDPOINT/api/v2/connectorV2/mcpservers/jira" \
  -d '{
    "properties": {
      "description": "Jira connector",
      "state": "Enabled",
      "connectors": [{
        "name": "jira",
        "connectionName": "jira",
        "displayName": "Jira",
        "operations": [
          {
            "name": "GetIssue",
            "displayName": "Get Issue",
            "agentParameters": [
              { "name": "issueId", "schema": { "type": "string", "required": true } }
            ]
          },
          {
            "name": "AddComment",
            "displayName": "Add Comment",
            "requiresApproval": true,
            "userParameters": [
              { "name": "projectKey", "value": "OPS" }
            ],
            "agentParameters": [
              { "name": "issueId", "schema": { "type": "string", "required": true } },
              { "name": "body", "schema": { "type": "string", "required": true } }
            ]
          }
        ]
      }]
    }
  }'

In this example:

  • GetIssue is enabled with no approval required.
  • AddComment requires approval (requiresApproval: true) and has projectKey locked to OPS.

For the full list of endpoints, see API reference.

Troubleshooting

Issue Cause Fix
OAuth popup doesn't appear Browser popup blocker Allow popups from sre.azure.com
"Sign in" button stays disabled Permission check failed Verify you have the SRE Agent Author or Administrator role on the agent (see User roles)
Connector not visible in picker Feature not enabled Confirm managed connectors preview is enabled for your agent
Agent doesn't use the connector Operations not selected Edit the connector and verify at least one operation is enabled
Dynamic dropdown loads slowly Service is responding slowly Wait a moment or type the value manually

Related content

  • Last updated on