Gemini CLI updates: Custom commands, OAuth, Vim mode, and more

Smarter MCP Clients: Making LLM Tooling Leaner and Faster At GoDaddy, we’ve been exploring ways to make large-scale LLM integrations more efficient — not just for us, but for the entire ecosystem. Today, most Model Context Protocol (MCP) clients inject every tool definition into every prompt. That’s fine for a few tools — but at scale, it means wasted tokens, slower responses, and higher compute costs. Our proposal — Advertise-and-Activate — offers a simple alternative: let clients advertise tool summaries first, then activate full definitions only when needed. This approach can:  ✅ Cut token usage by up to 94%  ✅ Reduce latency and context bloat  ✅ Improve LLM reasoning and scalability  ✅ Require no protocol changes — it works today  It’s a small architectural shift with a big impact on performance and cost. We’re already testing it internally across engineering workflows — from infrastructure management to developer tooling — and the results are promising. This is a key piece of our broader vision for a scalable AI agent ecosystem. While our Agent Name Service (ANS) implementation focuses on secure, internet-scale discovery and verification, patterns like Advertise-and-Activate are crucial for making subsequent agent-to-agent communication via protocols like MCP both efficient and enterprise-ready. 📄 Read the full article here: Smarter MCP Clients: A Proposal for Advertise-and-Activate https://lnkd.in/gAG7gmxr We’d love to hear your thoughts or experiences. — Jay Gowdy & Travis Muhlestein #AI #LLM #MCP #GoDaddy #Engineering #MLOps #SystemsDesign #Scalability 

Always proud of our teams pushing the boundaries of AI efficiency and scalability at GoDaddy. The work below highlights a key idea we’re exploring — a simple architectural shift that could significantly reduce compute costs and latency in large-scale LLM integrations, benefiting not just us but the wider ecosystem. Excited to see innovations like Advertise-and-Activate helping shape a more scalable agent infrastructure for the future. 👇 More details in the note below.

Exciting news for the developer community: the next version of the Model Context Protocol is set to be released on November 25th, 2025. This upcoming release will introduce crucial improvements, including support for asynchronous operations, enhanced statelessness for better scalability, and standardized server identity. These updates will strengthen MCP as the backbone for production-level AI integrations. For a detailed breakdown of the timeline and priority areas, read the full announcement: https://lnkd.in/gwACPEgX #ModelContextProtocol #MCP #OpenSource #ArtificialIntelligence #Roadmap

I recently had the chance to sit down with Tori Seidenstein, Co-Founder & CEO of Tadata, and came away genuinely excited about what her team is building. Tadata is tackling one of the biggest challenges in AI infrastructure right now: making APIs instantly usable by AI agents through the Model Context Protocol (MCP), which is quickly becoming the new standard for AI-tool connectivity. Here’s what really stood out: • Over 2,000 organizations already using Tadata • Their open-source FastAPI-MCP library has 10.7k GitHub stars • Trusted by major enterprises like Microsoft, Cisco, and Neo4j • The MCP server market is projected to reach $5.5B by 2034 What makes Tadata special: • Just 3 lines of code: Turn any FastAPI app into an MCP server instantly • Enterprise-grade security: Built-in authentication with your existing FastAPI setup • Zero maintenance: Their hosted platform (tadata.com) removes infrastructure overhead • Production-ready: Keeps your schemas, docs, and deployment options intact • Developer-first: Works smoothly with FastAPI, Flask, and raw OpenAPI specs Why this matters: as OpenAI, Copilot, Claude, and Cursor all move toward MCP, companies need an easy, secure bridge between their existing APIs and this new AI ecosystem. Tadata isn’t just building that bridge, they’re making it effortless, reliable, and ready for the enterprise. What I love most is that they treat MCP as a natural extension of your API framework, not a bolt-on layer. That means your existing security, logic, and structure stay exactly as they are, while AI agents gain structured access in a way that actually scales. From modernizing enterprise systems to helping startups move faster, Tadata is perfectly positioned at the crossroads of the API economy and the AI revolution. I have no doubt that Tori and her team will play a major role in accelerating MCP adoption and shaping how developers connect their APIs to the next generation of AI systems!

🚀 Just Built My First Remote MCP Server! What I learned about Model Context Protocol (MCP) and why it's a game-changer for AI tools: The Challenge: Claude Desktop expects local MCP servers via stdio, but I wanted to deploy remotely via HTTP. The Solution: Built a protocol bridge that translates between stdio and HTTP - basically making a remote server appear local to Claude Desktop. https://zurl.co/4wpj7 Key Technical Insights: Streamable HTTP > WebSockets for MCP - works with existing infrastructure (load balancers, CDNs, firewalls) Bridge Pattern is Essential - You need a translator between Claude's stdio expectations and your HTTP server JSON-RPC over HTTP - Clean protocol that's easy to debug and scale Error Filtering Matters - Claude Desktop validates JSON schemas strictly, so notification errors with null IDs need filtering Architecture That Actually Works: Claude Desktop ←→ Python Bridge ←→ Remote HTTP Server (stdio) (translator) (Railway/Heroku) What I Built: Task management MCP server deployed on Railway FastAPI + Streamable HTTP transport Full CRUD operations (create, search, complete tasks) Works with Claude Desktop, testable with MCP Inspector Could easily adapt for ChatGPT with REST wrapper The "Aha!" Moment: MCP isn't just about local tools - it's about making any service AI-accessible with proper protocol translation. The bridge pattern unlocks remote deployment while maintaining MCP's elegant design. Next up: Planning to build MCP servers for CRM integration, code analysis, and team workflows. Anyone else experimenting with remote MCP deployments? Would love to hear about different transport approaches! #AI #MCP #ClaudeDesktop #FastAPI #RemoteServers #AITools #TechStack

🚀 Just Built My First Remote MCP Server! What I learned about Model Context Protocol (MCP) and why it's a game-changer for AI tools: The Challenge: Claude Desktop expects local MCP servers via stdio, but I wanted to deploy remotely via HTTP. The Solution: Built a protocol bridge that translates between stdio and HTTP - basically making a remote server appear local to Claude Desktop. https://zurl.co/4wpj7 Key Technical Insights: Streamable HTTP > WebSockets for MCP - works with existing infrastructure (load balancers, CDNs, firewalls) Bridge Pattern is Essential - You need a translator between Claude's stdio expectations and your HTTP server JSON-RPC over HTTP - Clean protocol that's easy to debug and scale Error Filtering Matters - Claude Desktop validates JSON schemas strictly, so notification errors with null IDs need filtering Architecture That Actually Works: Claude Desktop ←→ Python Bridge ←→ Remote HTTP Server (stdio) (translator) (Railway/Heroku) What I Built: Task management MCP server deployed on Railway FastAPI + Streamable HTTP transport Full CRUD operations (create, search, complete tasks) Works with Claude Desktop, testable with MCP Inspector Could easily adapt for ChatGPT with REST wrapper The "Aha!" Moment: MCP isn't just about local tools - it's about making any service AI-accessible with proper protocol translation. The bridge pattern unlocks remote deployment while maintaining MCP's elegant design. Next up: Planning to build MCP servers for CRM integration, code analysis, and team workflows. Anyone else experimenting with remote MCP deployments? Would love to hear about different transport approaches! #AI #MCP #ClaudeDesktop #FastAPI #RemoteServers #AITools #TechStack

🚀 Just Built My First Remote MCP Server! What I learned about Model Context Protocol (MCP) and why it's a game-changer for AI tools: The Challenge: Claude Desktop expects local MCP servers via stdio, but I wanted to deploy remotely via HTTP. The Solution: Built a protocol bridge that translates between stdio and HTTP - basically making a remote server appear local to Claude Desktop. https://zurl.co/4wpj7 Key Technical Insights: Streamable HTTP > WebSockets for MCP - works with existing infrastructure (load balancers, CDNs, firewalls) Bridge Pattern is Essential - You need a translator between Claude's stdio expectations and your HTTP server JSON-RPC over HTTP - Clean protocol that's easy to debug and scale Error Filtering Matters - Claude Desktop validates JSON schemas strictly, so notification errors with null IDs need filtering Architecture That Actually Works: Claude Desktop ←→ Python Bridge ←→ Remote HTTP Server (stdio) (translator) (Railway/Heroku) What I Built: Task management MCP server deployed on Railway FastAPI + Streamable HTTP transport Full CRUD operations (create, search, complete tasks) Works with Claude Desktop, testable with MCP Inspector Could easily adapt for ChatGPT with REST wrapper The "Aha!" Moment: MCP isn't just about local tools - it's about making any service AI-accessible with proper protocol translation. The bridge pattern unlocks remote deployment while maintaining MCP's elegant design. Next up: Planning to build MCP servers for CRM integration, code analysis, and team workflows. Anyone else experimenting with remote MCP deployments? Would love to hear about different transport approaches! #AI #MCP #ClaudeDesktop #FastAPI #RemoteServers #AITools #TechStack

🚀 New article about MCP in Enterprise Just published my deep dive on Model Context Protocol (MCP) — exploring how this emerging standard can be applied inside enterprises: from developer sandboxes and local tools to corporate assistants, API gateways, and observability. In the piece, I break down: - why MCP matters beyond the hype, - what challenges appear in enterprise environments, - and which open-source tools (Anthropic MCP Registry, Docker MCP Catalog, IBM ContextForge, and others) can already help you get started. If you’re working with LLM integration, AI agents, or API modernization, this might be useful. 🧩 Read here: https://lnkd.in/e9CFSVGC Would love your thoughts or experiences with MCP in production — let’s discuss how this could shape the next layer of AI infrastructure. #MCP #AIInfrastructure #EnterpriseAI #OpenSource #LLMAgents #APIModernization

🚀 Just Built My First Remote MCP Server! What I learned about Model Context Protocol (MCP) and why it's a game-changer for AI tools: The Challenge: Claude Desktop expects local MCP servers via stdio, but I wanted to deploy remotely via HTTP. The Solution: Built a protocol bridge that translates between stdio and HTTP - basically making a remote server appear local to Claude Desktop. https://zurl.co/4wpj7 Key Technical Insights: Streamable HTTP > WebSockets for MCP - works with existing infrastructure (load balancers, CDNs, firewalls) Bridge Pattern is Essential - You need a translator between Claude's stdio expectations and your HTTP server JSON-RPC over HTTP - Clean protocol that's easy to debug and scale Error Filtering Matters - Claude Desktop validates JSON schemas strictly, so notification errors with null IDs need filtering Architecture That Actually Works: Claude Desktop ←→ Python Bridge ←→ Remote HTTP Server (stdio) (translator) (Railway/Heroku) What I Built: Task management MCP server deployed on Railway FastAPI + Streamable HTTP transport Full CRUD operations (create, search, complete tasks) Works with Claude Desktop, testable with MCP Inspector Could easily adapt for ChatGPT with REST wrapper The "Aha!" Moment: MCP isn't just about local tools - it's about making any service AI-accessible with proper protocol translation. The bridge pattern unlocks remote deployment while maintaining MCP's elegant design. Next up: Planning to build MCP servers for CRM integration, code analysis, and team workflows. Anyone else experimenting with remote MCP deployments? Would love to hear about different transport approaches! #AI #MCP #ClaudeDesktop #FastAPI #RemoteServers #AITools #TechStack

🚀 Just Built My First Remote MCP Server! What I learned about Model Context Protocol (MCP) and why it's a game-changer for AI tools: The Challenge: Claude Desktop expects local MCP servers via stdio, but I wanted to deploy remotely via HTTP. The Solution: Built a protocol bridge that translates between stdio and HTTP - basically making a remote server appear local to Claude Desktop. https://zurl.co/4wpj7 Key Technical Insights: Streamable HTTP > WebSockets for MCP - works with existing infrastructure (load balancers, CDNs, firewalls) Bridge Pattern is Essential - You need a translator between Claude's stdio expectations and your HTTP server JSON-RPC over HTTP - Clean protocol that's easy to debug and scale Error Filtering Matters - Claude Desktop validates JSON schemas strictly, so notification errors with null IDs need filtering Architecture That Actually Works: Claude Desktop ←→ Python Bridge ←→ Remote HTTP Server (stdio) (translator) (Railway/Heroku) What I Built: Task management MCP server deployed on Railway FastAPI + Streamable HTTP transport Full CRUD operations (create, search, complete tasks) Works with Claude Desktop, testable with MCP Inspector Could easily adapt for ChatGPT with REST wrapper The "Aha!" Moment: MCP isn't just about local tools - it's about making any service AI-accessible with proper protocol translation. The bridge pattern unlocks remote deployment while maintaining MCP's elegant design. Next up: Planning to build MCP servers for CRM integration, code analysis, and team workflows. Anyone else experimenting with remote MCP deployments? Would love to hear about different transport approaches! #AI #MCP #ClaudeDesktop #FastAPI #RemoteServers #AITools #TechStack