chatgpt-app-submission

To prepare an MCP server for ChatGPT Apps submission, validate tool annotations, CSPs, and outputSchema presence, then generate a compliant chatgpt-app-submission.json file. This process inspects repository metadata, README files, and manifests to ensure declared hints align with actual tool behavior across all MCP tools.

The chatgpt-app-submission.json file includes app_info, tools, and test_cases sections. It extracts app metadata, tool descriptors, resource templates, and widget metadata to provide reviewer-facing findings and outputSchema warnings for submission readiness.

Validating tool annotations and CSPs for ChatGPT Apps involves inspecting the MCP server codebase to check tool names, descriptions, and whether outputSchema is declared. The process surfaces review-check findings and requests approvals for needed source changes to ensure submission compliance.

ChatGPT Apps submission checks whether outputSchema is declared for MCP tools and reports warnings when it is missing. Ensuring alignment between declared tool hints and actual behavior across all MCP tools is required for the submission to be considered ready for review.

Generating test cases for an MCP server submission involves inspecting tool implementations and descriptors to produce both positive and negative test cases. These are included in the test_cases section of the chatgpt-app-submission.json file to demonstrate tool behavior for reviewers.

If MCP tool declarations do not match actual behavior, the submission validation identifies misalignment between declared hints and tool implementations. It surfaces reviewer-facing findings and requests the necessary source code approvals to correct the discrepancies before submission.