MCP Events and Plugin Creator Expand ChatGPT's Automation Platform

Two quieter DevDay 2026 launches carry much of the platform's load: OpenAI announced support for the proposed MCP Events specification, letting connected apps notify ChatGPT when something changes — a new task, message, or content update — and trigger automations, implemented by developers via webhooks. Alongside it, a conversational Plugin Creator, a simplified submission process with review tracking, and in-conversation plugin recommendations rebuilt the funnel through which plugins get made, approved, and discovered.

This piece explains what MCP Events adds to the Model Context Protocol picture, why event-driven triggering is the missing half of AI automation, how the creation-to-discovery pipeline changed, and what developers should build first. It is written for engineers integrating products with ChatGPT and teams automating on top of it.

MCP Events: what OpenAI is supporting

The Model Context Protocol standardized how AI applications call into tools — request-response, assistant-initiated. The proposed MCP Events specification addresses the reverse direction: the tool speaks first. Under it, a connected app can emit notifications when its state changes, and ChatGPT can receive them and fire automations in response. OpenAI's implementation path is explicit and practical — developers wire this up via webhooks; it is not automatic.

Backing a proposed open specification, rather than shipping a proprietary event channel, is itself a statement: OpenAI is betting that the MCP ecosystem — already the lingua franca for tool connectivity across the industry — should carry eventing too, with ChatGPT as its largest consumer.

Why events complete the automation story

Without events, AI automation is either human-initiated or clock-initiated — someone asks, or a schedule fires and the system polls for what changed. Both patterns leave the common case on the table: work that should begin the moment something happens. Event triggers close that gap, and DevDay's other launches show where the energy goes — shared tasks respond to events like emails and Slack messages, and dots pick up work as it appears.

For product developers, emitting MCP events is how your product becomes a first-class citizen of that loop. A project tracker that emits task-created events, a support desk emitting new-ticket events, a CMS emitting content-updated events — each becomes a source that ChatGPT-side automations, and users' agents, can react to instantly rather than on the next poll.

The rebuilt plugin funnel

Creation dropped to conversation level: the Plugin Creator walks a user through describing a workflow, adding instructions, and referencing files to produce a reusable plugin — no SDK required for the simple cases. Submission, the historical sore point, gained review tracking, requested-fix flows, and updates that no longer restart the whole process. Discovery gained the placement that matters most: ChatGPT now recommends relevant plugins inside conversations, alongside the directory.

The funnel now runs end to end: conversational creation lowers the top, tracked review unclogs the middle, and in-conversation recommendation gives the bottom real distribution. Combined with the same day's plugin extensions — full applications with sidebar homes and interactive panels — the plugin platform spans from prompt-built utilities to native apps.

Who should build what

Product teams with notification-rich systems should prioritize MCP Events support: being an event source positions your product inside users' automations, which is stickier than being one more callable tool. The webhook implementation cost is modest for anyone already running outbound notifications, and early spec adopters will shape it — it is proposed, not finalized.

Teams without engineering budget should mine the Plugin Creator: internal workflows described conversationally become reusable plugins, and the improved discovery means even narrow utilities can find their audience. For established plugin publishers, the immediate action is re-examining submission backlogs — the tracked-review process changes the cost of iterating on a live listing.

Spec adoption and event metering: what to watch

The specification's trajectory is the thing to track: whether MCP Events moves from proposed to adopted across the broader MCP ecosystem — other assistants, other runtimes — or remains ChatGPT-flavored in practice. Also unpublished: rate and delivery semantics for events at scale, and how event-triggered automations meter against plan usage.

The strategic read pairs with the rest of DevDay: extensions gave developers a surface inside ChatGPT, the Marketplace gave them a business channel, and MCP Events gives their products a voice — the ability to start conversations rather than wait to be called. Platforms win when integration runs in both directions; this is the second direction.