Using the MCP services at: awsnews.remotemcp.directory

Operations bound to the services offered via the following Streamable HTTP endpoint:
awsnews.remotemcp.directory

Chat UI Options:
* linkeddata.uriburner.com/chat
* inspect.mcp.garden#chat (via linkeddata.uriburner.com/chat/mcp/mes...)

One more thing, OPAL implements MCP as both a Client and Server, so it can bind to operations (tools) published by other MCP Servers that support SSE or Streamable HTTP.

Links in the next reply for a quick example.

🧰 This is what modern AI infrastructure looks like.

Here’s a quick showcase featuring AnthropicAI’s Claude Desktop interacting with a live OPAL instance that queries the DBpedia Knowledge Graph—entirely instructed using natural language.

#MCP #OPAL #Screencast

⚙️ Deploying to the cloud? It’s simple:

1. Go to the AWS Marketplace
2. Instantiate the OPAL AMI
3. Open {cname}/chat and log in
4. Plug in your favorite LLM API key from providers

Enjoy!

3/N 🧵

Another major benefit of Anthropic’s Claude Desktop support for remote MCP servers is the ability to combine powerful data access and visualization in notebooks that can be shared via web links.

OPAL MCP server example: an earlier demo ⤵️

claude.ai/share/11cb40...

What’s happening here?

An A2A Orchestrator is interacting with an LLM-powered AI Agent that exposes procedures, functions, 3rd party APIs etc., as MCP-accessible tools.

All made possible by the, in deceptively simple manner, by the OpenLink AI Layer (OPAL).

#MCP #A2A #GenAI

Do you have a link to a doc about how to setup Remote MCP Servers that support OAuth or Bearer Token based authentication?

Right now, you aren’t handling 401’s that trigger the OAuth handshake and I don’t know what you support regarding Authentication Headers.

What’s going on here?

Prompts are being processed through queries executed against a Knowledge Graph (KG)—all thanks to loose coupling. The target KG could reside locally, live on the Web, or be part of the vast Linked Open Data (LOD) cloud.

#A2A #MCP #VirtuosoRDBMS #AI #Agentic #GenAI

Architecturally, as depicted in the attached image.

BTW -- @Google Gemini can generate pretty accurate diagrams using @drawio XML!

#A2A #MCP #OPAL #VirtuosoRDBMS #AI #Agentic

It's been a while!

BTW -- I am working pretty extensively on MCP right now, so I would certainly like to have a play with Xogot.

Is there a repo link?

Continuation, as I see platform-specificity creeping into the notion of an MCP registry service.

github.com/modelcontext...

#MCP

Done.

github.com/modelcontext...

No, but a link-relation would be a neat addition to the spec since centralized registries will never solve the discovery problem.

The following should do.
<link rel="service" href="https://api.example.com/mcp">

I'll propose it :)

What a journey it’s been—starting with the ambitions of CORBA, the data connectivity standards of ODBC and JDBC, the global reach of the Web, and now, in the age of AI, it all converges through the Model Context Protocol (MCP).

Yes—if the MCP servers support Server-Sent Events (SSE) or Streamable HTTP, it’s possible. The main challenge right now is that most MCP clients are still STDIO-based, primarily due to lagging implementation of OAuth 2.1 for authentication.

Links:

1. github.com/OpenLinkSoft... -- ODBC

2. github.com/OpenLinkSoft... -- JDBC

3. github.com/OpenLinkSoft... -- pyODBC

4. github.com/OpenLinkSoft... -- ADO.NET

Very interesting stuff will happen, as demonstrated by the sheer number of MCP Servers. On our part, we offer MCPs for the following database connectivity protocols:

1. ODBC
2. JDBC
3. pyODBC
4. ADO.NET

Why?
To reintroduce DBMS independence in the age of AI :)

Yep!

DNS – Enables scalable machine name lookups across the Internet.

HTTP – Enables lookups of entities, their types, and their relationships at Web scale—where the Web serves as an abstraction layer over the Internet.

MCP – Enables access to tools (functions, operations) at Web scale.