Utsav Ladani

Enterprise MCP Architect. Core engineer behind Publishio. Bridging the gap between AI and enterprise security (RBAC, SIEM, Observability).

[Utsav Ladani @utsav_ladani]

@pamelafox I need a skill for copilot like code review, so that I can keep my PR clean.

[Utsav Ladani @utsav_ladani]

@harden_369741 Exactly, no need for complex context memory setup

[Utsav Ladani @utsav_ladani]

I end up having .<project-name> dir in my every project. It contains everything about the project that AI can't derive from the code. My decisions, coding taste, external context, backlog, preferences, etc.

This is the simplest way to give on-demand context to AI agent w/o bloating the good initial AI context. Try it for one of your project and see the difference.

[Utsav Ladani @utsav_ladani]

@jancurn Why didn't I know this! This is the perfect cli tool for my coding agent to verify MCP implementation I am building.

May be I could use it for writing automated tests for CI as well.

Thanks for sharing.

[Utsav Ladani @utsav_ladani]

@pietrozullo @rstagi_ Agree, these days frontier model companies mostly own them and evolving them good way, so custom one only makes sense in very privacy concerned firms.

[Utsav Ladani @utsav_ladani]

@berman66 Agree. Also, MCP is one of the vulnerable target in agent if not done correctly.

[Utsav Ladani @utsav_ladani]

@rstagi_ @pietrozullo Something similar, but at MCP host side https://t.co/su9mwS1a9a

[Utsav Ladani @utsav_ladani]

@dani_avila7 It would be more helpful if Claude and ChatGPT add this because it's needed at enterprise level.

[Utsav Ladani @utsav_ladani]

Great news for Claude Team & Enterprise plan users.

Finally Enterprise Managed Auth is in Claude MCP.

It means if a user is logged into your identity provider account, then they will be automatically logged in to MCP connectors if that identity provider is enabled for that service.

Ease for teams!

[ClaudeDevs @ClaudeDevs]

Users can now auto-connect plugins when they log in to Claude, with MCP servers that accept Claude as a trusted issuer.

If you’re building plugins for Claude, add support so users get started faster and stay connected without re-authorizing.

[Utsav Ladani @utsav_ladani]

tl;dr: this can make much difference if added. It's like getting benefit from both worlds.

If JS/TS to PHP transpiler would exist then it would provide great DX as well make WordPress more powerful for AI validations.

WordPress has all, Auth, RBAC, logs, everything someone need to build nice AI-native agentic CMS. Only missing piece is the content validation on server.

Currently, few things written in JS can only run in browser when block editor is open. For example, a static block.

But AI agents mostly call API which runs on server in PHP, so validation added in JS is ignored and people end up with broken page or markup. Believe me I see this case often.

Also, a reason for people who say PHP is dead to come back to WordPress - the best CMS. 😜

[Utsav Ladani @utsav_ladani]

Hey, I also built similar JS to PHP transpiler with limited syntax support. It worked and I am able to create a WordPress theme in React with SSR support.

My reason was that I want to build a system where I can just pull up @shadcn components and built pages like how I do in React.

But dropped midway because adding JS syntax and feature support was time taking.

I am really interested to talk to you about this idea....

[Utsav Ladani @utsav_ladani]

@sohmehta Very true. I can confidently built a complex feature in one prompt and can trust generated code that it won't have any bug.

It's go to Model for people who are running agents to write code and don't want to review it.

[Utsav Ladani @utsav_ladani]

@icanvardar That's good. I hate comments. Code should be self explanatory and comments should be used for method and function definitions and some hack made in the codebase.

Still, I feel AI is used to add comments on subsequent prompts when I challenge the AI decision.

[Utsav Ladani @utsav_ladani]

@KaranVaidya6 @composio Yeah, Specialized integration is a big thing. It needs agent SDKs which can quickly build them solutions.

[Utsav Ladani @utsav_ladani]

@swyx That's the way forward. Now I don't focus on learning new lang, just understand what a language can do what its specialities to verify the outcome and done

[Utsav Ladani @utsav_ladani]

@sohmehta New marketplace to get more users

[Utsav Ladani @utsav_ladani]

@pederzh Yay!

[Utsav Ladani @utsav_ladani]

@liadyosef @ndmccormack @OpenAI Is there any open conversation or any public issue or draft where I can keep an eye on it?

[Utsav Ladani @utsav_ladani]

@berman66 The granularity and simplicity to add restriction from hooks is much better compared to what we get at the gateway or runtime level.

[Utsav Ladani @utsav_ladani]

@ndmccormack @OpenAI Yes, because sometimes it's tricky for a user to give input in plain text, so having an interactive app can make a user's life easier.