What Connector Namespace is
Before GA, the service was documented as a fully managed integration service that hosts a catalog of prebuilt, reusable connectors and MCP servers that apps consume through a consistent programming model. The namespace is the Azure resource that does the heavy lifting. Per the preview-era post, it loads and runs operations, maintains connection state and credentials, polls source systems, dispatches webhook events, and applies retry and diagnostic policies.
Microsoft also says the service is separate from Logic Apps. Its documentation states that connector namespaces don't require, use, or change anything in Azure Logic Apps.
Compute support covers more than Functions. Azure App Service, Azure Container Apps and Azure Functions can all use connector operations, and any self-hosted compute can use connectors through a namespace.
The Functions launch that arrived with it
Microsoft's Ignite 2026 post says Managed Connectors in Azure Functions is now generally available together with Azure Connector Namespace. A companion Apps on Azure post says developers can:
- Connect to 1,600+ services through the managed connector ecosystem, including Microsoft 365, Salesforce, Dataverse, Jira, Slack and Dropbox.
- React to events with connector triggers.
- Call connected services through language-native connector SDKs.
- Build in C#, Python and TypeScript.
The SDK packages are Azure.Connectors.Sdk for .NET, azure-connectors for Python and @azure/connectors for Node.js.
The Ignite post says the integration targets .NET isolated, Python and Node.js. It describes the old pain as custom OAuth token refresh code, webhook registration and public inbound callback endpoints. The namespace is meant to take over authorized connections, token refresh and throttling.
Webhook versus polling
Polling is the main addition at GA. Microsoft says managed connector triggers in Functions previously relied on webhook delivery, and now you can choose either mode.
| Webhook | Polling (new at GA) | |
|---|---|---|
| Direction | The namespace pushes events to the function app over HTTPS | The function app pulls queued events outbound over HTTPS |
| Validation or identity | Entra ID or system keys | System-assigned or user-assigned managed identity |
| Public callback endpoint | Required | Not needed |
| Scaling | Per Microsoft, aimed at low-latency delivery | Scales on pending events |
Microsoft's example host.json values are defaultTargetPendingEventThreshold: 16 and maxPollingInterval: "00:00:30". These are examples, not tuned recommendations. Other tunable settings include a maximum batch size, because polling supports batching, so one invocation can process multiple events.
For locked-down environments, outbound-only polling avoids exposing an inbound endpoint. The trade-off is that it is pull-based. Webhooks suit workloads where latency matters most.
Developer tooling
For GA, Microsoft added a Connector Triggers experience to the Azure Functions extension for VS Code. In the Create New Project flow, a Connectors tab lets you pick a connector and event, scaffold the project, and be guided through creating or reusing the namespace, connection and trigger resources for local debugging. Azure Functions Core Tools also ships connector trigger templates. Sample repositories include Bicep templates, so azd up can provision and deploy them.
Caveats before you commit
- Typed SDK coverage is uneven. Strongly typed payloads and clients exist for popular connectors such as Outlook, SharePoint and Teams. Other connectors can still send events as JSON through the generic trigger. Microsoft warns that generic trigger support doesn't automatically mean SDK action support.
- Functions support has limits. Microsoft Learn's Functions overview, updated November 17, lists .NET 10 and 11 isolated, Python 3.13+ and Node.js 24+. It says Java, PowerShell and Go aren't supported. Hosting plans are Flex Consumption (recommended), Premium, Dedicated and Container Apps. Linux-hosted apps are supported in all regions, while Windows-hosted apps have limited availability.
- Enterprise connectors may lag. The preview documentation said high-usage and standard connectors come first, with enterprise connectors such as SAP, IBM MQ and Oracle Database following in later waves. I found no GA-era confirmation of that roadmap, so check the catalog for the connector you need.
- Some documentation hasn't caught up. Several Learn and blog pages I found still carry preview language. One says during preview, this capability is only available in Azure public regions, and another says there is no SLA during preview. These pages predate GA, so don't treat their limits as the GA terms. The Functions page I found in search still said the integration is in public preview. Confirm current region availability and service terms in the live documentation before you plan a rollout.
Pricing
Microsoft's GA post says the Functions integration carries no extra charge of its own. Two components apply:
- Standard Azure Functions pricing, based on your hosting plan.
- Connector usage, billed per connector call. Rates depend on whether the connector is classified as Standard or Enterprise, as set out in the Connector Namespace pricing.
Microsoft Learn adds that Connector Namespace has separate consumption-based pricing. I haven't verified actual rates.
Trying it out
Microsoft's quickstart describes a portal flow. These steps come from the pre-GA quickstart, so menu labels may change.
- In the Azure portal, search for "connector namespace" and select it.
- Select Create. Choose a subscription, resource group, a unique namespace name (2–64 characters: alphanumerics, hyphens or underscores) and a region.
- Review identity. A system-assigned managed identity is on by default, and you can add user-assigned identities.
- After deployment, open the namespace. The portal redirects you to the Connector Namespaces portal, where you sign in with Microsoft.
- Create connections first. Triggers need them. Then add triggers, or MCP servers that expose connectors as tools for AI agents.
The portal isn't the only route. Microsoft's documentation also lists ARM and Bicep templates and the Azure CLI, although a third-party preview write-up said Bicep support wasn't yet available. That may have changed by GA.
Where it fits
Microsoft's own guidance is that managed connectors in Functions are additive. For pure orchestration across connectors with no custom code, it says Logic Apps Standard remains the simplest choice. Functions with connectors suits code-first work such as custom branching, other bindings and AI model calls between trigger and action. App Service with connectors suits adding connector events to an existing web app or API.
The MCP angle matters for AI agent work. The namespace can host MCP servers, which spares teams from running their own. Preview-era documentation listed Playwright and Azure SQL as hosted servers, and Microsoft says the catalog expands over time. I haven't verified the GA catalog.
The GA stamp does not mean every connector, trigger and region is covered equally. Treat the catalog and the Functions support matrix as the source of truth, and pilot with the one or two connectors you actually need.
References
- (Launched) Generally Available: Azure Connector Namespace Azure Updates · 2026-11-17T20:00:32Z
- Create and manage connector namespaces for integration - Connector Namespace | Microsoft Learn learn.microsoft.com
- Managed connectors in Azure Functions are now generally available | Microsoft Community Hub techcommunity.microsoft.com