Blog

TMS AI Studio v1.4 is bringing HTTP.sys to MCP

Today

TMS AI Studio keeps evolving, and in the new v1.4.0.0 release we’re introducing a major upgrade to the HTTP-based MCP transport layers. On Windows, both TTMSMCPSSETransport and TTMSMCPStreamableHTTPTransport now run on HTTP.sys — Microsoft’s kernel-level HTTP server stack.

The benefits of switching to HTTP.sys on Windows is clear:

✅ Native performance: Requests are handled directly by the Windows kernel.

✅ Lightweight: No dependencies on 3rd party libraries.

✅ Secure: Reuse certificates from the certificate store instead of relying separate files.

This isn’t just a swap of the initial implementation. Under the hood we’ve created an abstract HTTP server layer, which in the future will allow us to plug in different engines per platform!

Changes to Transport Layers

This upgrade comes with a few adjustments, which were necessary in order to provide a more flexible and robust foundation for future extensions.

❌ Removed SSLVersion: There’s no need to pick TLS versions anymore. TLS 1.0 and 1.1 are deprecated by many browsers and is now also disabled in Windows by default.

🔁 Updated Custom Header Events: The OnCustomHeaderCheck and OnCustomHeaderSet events have been replaced by OnGetCustomHeader and OnSetCustomHeader to enable using the abstraction layer.

🔐 SSL Certificates: HTTP.sys does not rely on certificate files, so we are introducing a file-based system for the initial setup to remain consistent across platforms. If you are already using SSL, please refer to our documentation on how to get everything up and running again.

Real-World Use: ChatGPT and Claude Web Interfaces

To put the new architecture to the test, we connected one of our MCP servers directly to the ChatGPT and Claude web interfaces using their custom connector support. Our StreamableHTTP demo was the perfect fit, as it already exposes a streamable HTTP transport for remote access. For those unfamiliar, this demo server offers a single tool: fetching the current weather for a given city.

After building and deploying the server, we added it as a connector and used it directly inside a conversation. In ChatGPT, this requires enabling Developer Mode first, so follow the official guide if you haven’t done that yet.

Once the connector is active, we can start asking questions that the newly registered tool can answer:

Claude is just as easy to setup. You can follow this official guide to get up and running with custom connectors.

If you would like to repeat this with you own MCP server, be aware of the following requirements:

ChatGPT	Claude
Needs a Plus or Pro individual plan. For Business and Enterprise members this feature is being currently rolled out, check with the administrator to have this feature enabled. Requires HTTPS	Needs a Pro or Max individual plan. Team and Enterprise plans are also supported, but only the owner can add new connectors. Requires HTTPS Cannot connect to a connector with a port in it's address. For example: https://myserver.com:1234/mcp won't work.

Don't have access to paid ChatGPT or Claude but still want to see the server live? We decided to leave our demo application online as a showcase of what TMS AI Studio can do! Pick your own favourite MCP client application and use the following address to connect to our live MCP demo server: https://tmswebacademy.com/mcp. 

Conclusion

Switching to HTTP.sys is an important structural step and gives us a cleaner foundation for future improvements across all platforms. Testing against platforms such as ChatGPT and Claude confirms that TMS AI Studio provides a robust, practical MCP implementation that’s ready to be used in many different contexts.

The v1.4.0.0 release has more to offer! We’ll introduce the next addition in an upcoming post so stay tuned!

Tunde Keller

Related Blog Posts

1. Add AI superpower to your Delphi & C++Builder apps part 1


2. Add AI superpower to your Delphi & C++Builder apps part 2: function calling


3. Add AI superpower to your Delphi & C++Builder apps part 3: multimodal LLM use


4. Add AI superpower to your Delphi & C++Builder apps part 4: create MCP servers


5. Add AI superpower to your Delphi & C++Builder apps part 5: create your MCP client


6. Add AI superpower to your Delphi & C++Builder apps part 6: RAG


7. Introducing TMS AI Studio: Your Complete AI Development Toolkit for Delphi


8. Automatic invoice data extraction in Delphi apps via AI


9. AI based scheduling in classic Delphi desktop apps


10. Voice-Controlled Maps in Delphi with TMS AI Studio + OpenAI TTS/STT


11. Creating an n8n Workflow to use a Logging MCP Server


12. Supercharging Delphi Apps with TMS AI Studio v1.2 Toolsets: Fine-Grained AI Function Control


13. AI-powered HTML Reports with Embedded Browser Visualization


14. Additional audio transcribing support in TMS AI Studio v1.2.3.0 and more ...


15. Introducing Attributes Support for MCP Servers in Delphi


16. Using AI Services securely in TMS AI Studio


17. TMS AI Studio v1.4 is bringing HTTP.sys to MCP

This blog post has not received any comments yet.