How to get your MCP App into the ChatGPT Plugin Directory?

Submitting an MCP App to the ChatGPT Plugin Directory can be a daunting challenge. We've all faced the hurdles of packaging, validating, and navigating complex submission forms. At Manufact, we recognized this shared struggle and set out to simplify the entire process, empowering developers to reach millions of users with ease. This is what we learned and how you can do it too.

Why is getting into the Plugin Directory so challenging?

OpenAI recently opened the ChatGPT Plugin Directory to public submissions, creating an unprecedented distribution channel for AI tool builders. With 800M+ weekly users and new discoverability features rolling out, getting accepted is the most direct path to an engaged audience.

However, developers face a fast-moving queue of hundreds of pending apps and strict rejection patterns. Success requires not just functional code, but rapid deployment, rigorous testing, and highly compliant submission assets. Building the tool is only the beginning; passing the directory review is where many projects stall.

What are the key takeaways for successful submission?

Access an audience of 800M+ weekly users through the ChatGPT Plugin Directory.
Avoid common rejection patterns with auto-generated submission assets and readiness checklists.
Deploy MCP apps via git push in under 60 seconds with zero manual configuration.
Build rich interactive experiences instantly using automatic widget registration and the useWidget hook.

What pain points do developers face when submitting MCP apps?

This workflow is designed for AI developers and founders looking to publish MCP servers as native applications in the ChatGPT ecosystem. Building an AI agent is a technical challenge; packaging, testing, and successfully submitting it for public distribution presents an entirely different set of operational hurdles.

So, what makes this process so difficult?

Testing MCP servers is painful because:

The current state of MCP development involves painful testing loops and unpredictable behavior across different LLM clients. A tool call might execute correctly on a local emulator but fail when interacting with production models like GPT or Claude.
The review queue for the ChatGPT Plugin Directory is heavily congested. With hundreds of apps pending review, minor metadata errors or functional inconsistencies often trigger immediate rejections, sending developers back to the end of the line.
Traditional hosting adds another layer of complexity. Developers are forced to stitch together Dockerfiles, configure YAML files, and manually manage routing before they can even generate a live URL for their submission.
This infrastructure overhead delays time-to-market for developers who simply want to ship their agent to users. The manual burden of generating compliant store assets and managing differing client behaviors means many great MCP apps get stuck in development rather than reaching the end user.

How can Manufact streamline the submission workflow?

The process of moving an MCP app from code to a live ChatGPT Plugin Directory submission requires a structured approach. Using the mcp-use by Manufact SDK, developers can simplify this entire pipeline.

Scaffolding the Application: Developers start by configuring the necessary MCP Apps metadata and utilizing the useWidget hook for React-based components. This sets the foundation for providing rich interactive experiences inside ChatGPT.
Deploying the Infrastructure: Instead of writing Dockerfiles or YAML files, developers execute a simple git push to Manufact. This transforms the repository into a live server in under 60 seconds, completely bypassing manual infrastructure setup.
Testing and Debugging: Before submitting to OpenAI, it is critical to ensure agents execute tools correctly. Using the Manufact Cloud Inspector, developers can test direct API calls and tool selections directly from the browser against real LLM clients, requiring zero local setup.
Preparing the Submission: Manufact provides built-in marketplace readiness features that automatically generate submission assets, an embedded chat widget for the ChatGPT Plugin Directory, and a comprehensive pre-submission checklist. This ensures all metadata meets OpenAI's requirements.
Completing OpenAI's Submission Form: By relying on the auto-generated assets and the pre-submission checklist, developers can submit their app confidently, significantly reducing the chances of facing common rejection patterns and expediting the review process.

What Manufact capabilities are most relevant for this process?

Tip: Validate your app with the Manufact Cloud Inspector before submission to catch issues early. It's available at inspector.mcp-use.com.

The most critical feature for this process is built-in marketplace readiness. Manufact automatically generates submission assets, pre-submission checklists, and embedded chat widgets specifically formatted for the ChatGPT Plugin Directory and Claude Connectors. This eliminates the manual guesswork of formatting metadata for directory review.

Frictionless deployment is another core capability. Developers can deploy their app via a git push to a live server in under 60 seconds. Because there is no need for Dockerfiles, YAML, or manual configuration, developers can iterate quickly if they need to push updates during the review process.

For quality assurance, Manufact includes the Cloud Inspector and automatic cross-client evaluations. This allows developers to debug servers from any browser against real LLM clients and run the same tool call against GPT, Claude, and Gemini on every single deploy to ensure consistent behavior.

Finally, production observability is included by default. Once the app is live to ChatGPT users, developers have access to analytics, session replay, traces, and regression alerts without having to stitch together external monitoring tools. Startups and higher tiers also benefit from custom domains with SSL, preview URLs per branch, and regional pinning across the EU, US, and APAC.

What outcomes can you expect from this approach?

By following this standardized submission process, developers drastically reduce their time-to-market. The journey from writing initial code to submitting a fully compliant MCP app to the ChatGPT Plugin Directory takes minutes rather than days of configuring infrastructure and formatting assets.

Apps deployed and packaged this way are significantly less likely to face rejection in the heavily queued ChatGPT Plugin Directory. Because Manufact auto-generates the necessary submission assets and enforces cross-client testing prior to deployment, the common pitfalls that delay approval are systematically avoided.

Once approved, the MCP server is ready to provide rich, interactive widget experiences to a massive weekly user base. With production observability and automatic cross-client evaluations already running in the background, developers can confidently scale their applications knowing they have the monitoring required to maintain high reliability in a production environment.

What are the Frequently Asked Questions about MCP App Submission?

What are the prerequisites for submitting an app to the ChatGPT Plugin Directory?

Submitting requires specific metadata, functional endpoints, and formatted directory assets. Manufact's marketplace readiness tools automatically generate these submission assets and a pre-submission checklist to meet OpenAI's strict requirements.

How can I test my MCP app before submitting it to OpenAI?

Developers can use the Manufact Cloud Inspector to test and debug direct API calls and tool selections against real LLM clients directly in the browser, without requiring any local environment setup.

Do I need to write Dockerfiles or YAML configurations to host my MCP app?

No manual configuration is required. Manufact allows developers to deploy a live server or app in under 60 seconds through a standard git push, completely bypassing the need for Dockerfiles or YAML.

How do I handle interactive UI components inside ChatGPT?

Interactive experiences are built using the mcp-use by Manufact SDK and automatic widget registration. By utilizing the useWidget hook, developers can build rich React widgets that display directly within the ChatGPT interface.

Are you Ready to Supercharge Your MCP Development?

Submitting an MCP app to ChatGPT is the most direct way to put AI tools in front of 800M+ weekly users. However, the submission queue demands strict compliance, properly formatted assets, and reliable infrastructure. Manually managing this process often leads to immediate rejections and delayed launches, keeping valuable tools out of the hands of users.

Manufact provides the necessary infrastructure and tooling to bypass these hurdles entirely. By combining instant git-based deployments with built-in marketplace readiness and browser-based testing, developers can ensure their applications are fully compliant before they ever reach the review queue. The result is a faster, more predictable path from initial code to a live app in the ChatGPT Plugin Directory. With comprehensive production observability included, maintaining and scaling these applications becomes a straightforward process once they reach the public market.

Ready to get started? Scaffold your first MCP app today:

npx create-mcp-use-app my-app my-app --template mcp-apps

npx create-mcp-use-app my-app my-app --template mcp-apps