MCPKit is a blueprint for building authenticated Model Context Protocol (MCP) servers that let you bring proprietary data, content, and systems into ChatGPT, via ChatGPT Dev Mode.
Built by MCP engineers for enterprise builders, MCPKit accelerates the path from prototype to production when native connectors or existing MCP services do not cover your use case.
- Serve high-value data directly inside ChatGPT through the connectors platform while respecting your existing authentication and entitlement rules.
- Keep sensitive content behind an authorization layer you control; MCPKit uses Auth0 as the example but supports any OIDC-compliant provider.
- Financial services: Expose internal research reports, expert-call transcripts, and other alternative data. The synthetic bundle in this repo mirrors alternative data feeds you can adapt to production.
- Customer support & success: Expose a gated knowledge base that blends CRM data, playbooks, and ticket summaries.
- Healthcare & life sciences: Expose documentation, SOPs, and clinical trial data.
- E-commerce: Expose item inventory, order status, fulfillment and delivery statuses.
Use synthetic_financial_data/ as a realistic sandbox for pipelines and demos. It contains alternative data artifacts such as analyst reports, expert-call summaries, and web-search trends with consistent tickers and timestamps. Swap in your own feeds once you are ready to plug into live systems.
- Authenticated MCP server scaffolds: Python and TypeScript servers that implement a number of different tools, include Deep Research-compatible
searchandfetchtools, apply entitlement checks, and follow the recommended resource/authorization separation model. - Authorization patterns: The servers also implement the MCP authorization specification pattern of separate resource and authorization servers. For the authorization server, we use an end-to-end Auth0 integration that you can replace with Okta, Azure AD, or an internal IdP by updating JWKS resolution, token validation, and tenant metadata.
- Sample data: A complete
synthetic_financial_data/bundle with expert-call transcripts, pricing snapshots, and enrichment metadata so you can demo entitlement-aware tools before wiring up production feeds.
Both implementations share a consistent API surface, emit structured logs, and lean on Auth0 for token exchange. Replace Auth0 with your preferred authorization provider by updating the server configuration documented in each README.
- Run locally: Start either scaffold (
npm run devorpython -m server.app) with your Auth0 application or alternate authorization server settings. - Expose securely via ngrok: Tunnel the MCP server (
ngrok http <port>) so ChatGPT can reach it during development without production deployment. - Register in ChatGPT Dev Mode: Provide the tunneled URL, login with OAuth, and start querying.
- Harden for production: When you are ready, deploy your MCP server on your hosting platform of choice.