Uh oh!
There was an error while loading. Please reload this page.
feat(mcp): teach wizard modern capture support - #372
Conversation
Detect both MCP SDK majors in TypeScript and Python, enable conversation correlation for the handshake-free revision, and configure TypeScript model self-reporting where supported. Keep custom dispatchers protocol-aware and prevent the wizard from promising automatic resource, prompt, or Python model capture. Verified with npm test (173 tests) and npm run build.
🧙 Wizard CIRun the Wizard CI and test your changes against wizard-workbench example apps by replying with a GitHub comment using one of the following commands: Test all apps:
Test all apps in a directory:
Test an individual app:
Show more apps
Test against a wizard branch:
Add Results will be posted here when complete. |
lucasheriques
commented
Sep 1, 2026
/wizard-ci mcp-analytics |
🧙 Wizard CI ResultsTrigger ID:
Configuration
|
lucasheriques
commented
Sep 1, 2026
/wizard-ci mcp-analytics wizard:codex/mcp-analytics-ci-region |
🧙 Wizard CI ResultsTrigger ID:
Configuration
|
lucasheriques
commented
Sep 1, 2026
/wizard-ci mcp-analytics wizard:codex/mcp-analytics-ci-region |
🧙 Wizard CI ResultsTrigger ID:
Configuration
|
Require custom dispatchers to propagate the actual protocol revision on each capture and validate final report event names against the event reference.\n\nVerification:\n- npm test (173 tests)\n- npm run build
lucasheriques
commented
Sep 1, 2026
@PostHog/mcp-analytics could you review the modern MCP spec, model capture, and custom-dispatcher guidance? The pinned Wizard CI run scored 5/5 and 4/5 across the two fixtures; the follow-up commit addresses the two manual-review gaps. |
Summary
MCP server authors get modern-spec-aware Wizard guidance for TypeScript and Python instrumentation.
2026-07-28wrapping paths.posthog>=7.40.0when Python projects need MCP SDK 2.x support.Validation
npm test- 173 tests passed.npm run build- passed.Agent context
Codex compared the skill against the shipped TypeScript and Python SDK behavior, exercised both workbench fixtures, and reviewed their generated code and reports.