WHAT WE BUILD TODAY
A free Worksmarto MCP server that lets Claude, ChatGPT, Cursor and VS Code search the blog, fetch the newest posts and read today’s AI news, plus a setup pack to download.
Level: Beginner Stack: AI coding · WordPress REST API · Next.js · mcp-handler · GitHub · Vercel
Worksmarto now lives inside your AI assistant. One link, and Claude or ChatGPT can search every Worksmarto article, pull the post published five minutes ago and read the daily AI news, all without leaving the chat.
No sign-in. No API key. No cost. Just a URL.
THE 30-SECOND TAKE
- The Worksmarto MCP has three read-only tools: search the blog, get the latest posts, and get today’s AI news.
- It was built with AI coding on free APIs that already exist, mainly the WordPress REST API, and runs keyless on a free hosting plan.
- Any blog or micro SaaS can do the same, so AI assistants find new content without waiting for Google to crawl it.
Read it your way: see the whole idea in the picture below, then follow the five parts in order.
THE WORKSMARTO MCP
ONE URL. YOUR AI READS WORKSMARTO.
Free. Keyless. Read-only. Just add one URL.
worksmarto.com
YOUR LEARNING PATH
1. Everything in the Worksmarto Innovation Lab
First, the full shelf. Every product below is free to try, and the MCP is the newest one. The Innovation Lab is where each experiment lands once it works.
| Product | What it does |
|---|---|
| Worksmarto MCP · New | Puts Worksmarto inside Claude, ChatGPT, Cursor and VS Code. Try it below |
| Worksmarto AI News | The day in AI in two minutes, with a takeaway per story. news.worksmarto.com |
| Idea Wheel | Spin for an AI freelance service or solo product idea, with tools and a launch workflow. Spin the wheel |
| AI Workflow Architect | Maps AI workflows for tasks, clients and projects. architect.worksmarto.com |
| AI Agent Team | A team of AI agents that plan, vote, draft and review together. ai-agent-team.worksmarto.com |
Want every product, feature and post in one place? The Explore hub filters them by type and audience.
In short: five free tools, one shelf, and the MCP is the one that brings the rest of Worksmarto into your chat.
2. Try the Worksmarto MCP: One Link and a Download
Everything starts with one server URL. Copy it, paste it into your AI app, done.
Windows, Mac and Linux · Claude Desktop, Cursor and VS Code configs · sample SKILL.md · 4 KB
What is a custom MCP?
A custom MCP, called a custom connector in Claude, is any MCP server you add yourself by pasting its URL. The big apps come with built-in connectors like Google Drive or GitHub. Custom connectors let you plug in anything else, such as a blog, a store or your own micro SaaS. New to the idea? Start with What Is an MCP Server?.
How to use it: pick your app
Claude (web, desktop or phone): easiest
Settings → Connectors → Add custom connector. Name: Worksmarto. URL: the link above. Click Add. Then in any chat tap + → Connectors → turn on Worksmarto. The free plan allows one custom connector.
Claude Desktop on Windows: config file
Optional, needs Node.js. Press Win + R, paste %APPDATA%\Claude and press Enter. Open claude_desktop_config.json and paste the file of the same name from the setup pack. Fully quit and reopen Claude.
Cursor
Open C:\Users\you\.cursor\mcp.json (Mac: ~/.cursor/mcp.json), paste cursor-mcp.json from the pack, and restart Cursor.
VS Code with Copilot
Create .vscode/mcp.json in your project and paste vscode-mcp.json from the pack. Copilot agent mode picks it up.
ChatGPT
On paid plans, turn on developer mode under Settings → Apps & Connectors → Advanced, create a connector and paste the URL.
Try these prompts
Each answer comes back with Worksmarto links, so the next read is one tap away.
In short: copy one URL, paste it into Claude or ChatGPT, or download the pack for Cursor, VS Code and Claude Desktop on Windows.
3. Don’t Confuse MCP With SKILL.md
Both words show up in every AI tool now, and they do very different jobs.
| MCP | SKILL.md |
|---|---|
| A server the AI connects to | A text file of instructions the AI reads |
| Gives live data and actions: search, fetch, post | Gives know-how: steps, rules, a house style |
| Lives on the internet behind a URL | Lives in a folder you upload to the AI app |
| The hands | The habit |
The rule: an MCP lets the AI do something. A skill tells the AI how to do it well. The best setups use both.
Here is a copyable SKILL.md that teaches Claude how to use the Worksmarto MCP. It is also in the setup pack, inside the worksmarto-research folder.
To add it in Claude, zip the worksmarto-research folder and upload it under Settings → Capabilities → Skills.
In short: MCP is the connection, SKILL.md is the playbook, and the sample above shows both working together.
4. How We Built It
Why build an MCP at all?
Google does not crawl a new blog post the moment it goes live. For a young site it can take days, sometimes longer. During that gap, nobody searching on Google sees the post.
AI assistants are the other front door. With an MCP, an assistant asks Worksmarto directly and gets the newest post the same minute it is published. No crawl wait. No ranking needed. The MCP in 2026 post explains why this protocol is spreading so fast.
The build, step by step
1. AI coding: yes, you read that right
The whole server was written with AI coding in one chat: the tools, the tests, the landing page and the README. If you run a blog or micro SaaS, try this. Describe what you want, review what comes back, and ship.
↓
2. Three features, added carefully
Search ranks the best match first, not just the newest. Latest posts fetches what was just published. AI news reads the daily brief from Worksmarto AI News. Each tool was tested one by one before it went live.
↓
3. Keyless and read-only
No AI API key on the server. Worksmarto serves the data, and the user’s own Claude or ChatGPT writes the answer. Every tool is marked read-only, and nothing users send is stored.
↓
4. Push and deploy
The code went to GitHub, and Vercel deploys it automatically on every push, on the free Hobby plan.
An MCP is built on APIs
An MCP server is a thin, friendly layer on top of APIs that already exist. The AI talks MCP, the server talks API. Here is everything Worksmarto used, and all of it is free.
| Piece | What it does |
|---|---|
| WordPress REST API | Built into every WordPress site at /wp-json/wp/v2/posts. Free, public, no key. Powers search and latest posts. |
| Worksmarto AI News | The daily news page at news.worksmarto.com. Powers the AI news tool. |
| AI coding assistant | Wrote and tested the server code from plain-English instructions. |
| Next.js + mcp-handler | Turns a normal web project into an MCP endpoint at /mcp. |
| TypeScript + Zod | Checks every tool input so bad requests are rejected cleanly. |
| GitHub | Stores the code and triggers every deploy. |
| Vercel Hobby | Free hosting with HTTPS, which MCP apps require. |
| UTM tags | Every link carries utm_source=mcp, so MCP visits show up separately in analytics. |
Total running cost: ₹0. Curious how these layers fit a bigger product? See What Is an AI Full Stack Developer?.
In short: AI coding plus the free WordPress REST API gave Worksmarto a keyless, ₹0 MCP, and new posts reach AI assistants without waiting for Google.
5. How to Get Listed in the Claude and ChatGPT Directories
A custom connector works today for anyone with the URL. A directory listing goes further: people find the MCP inside the app and connect in one click. Here is what each directory asks for.
Claude Connectors Directory
Submit through the developer portal at claude.ai/directory/manage and choose MCP connector. Any paid Claude plan can submit. You need a public HTTPS server, a title plus a read-only or destructive label on every tool, a documentation URL, a privacy policy URL, a support contact and an icon. Approved servers are listed as Community connectors.
ChatGPT apps
Submit through the OpenAI platform. You need a stable HTTPS server, website, support, privacy and terms URLs, domain verification, organisation verification to publish under a company name, plus test cases and a demo video for review.
Before you submit
Test every tool as a custom connector first. Write a public docs page, like this post. Keep the privacy policy up to date. Worksmarto’s tools already carry read-only labels, and the directory submission is the next step.
Full details are in Anthropic’s connector submission guide and OpenAI’s app submission guide.
In short: test as a custom connector, publish a docs page and privacy policy, label every tool, then submit to each directory.
WorkSmarto verdict
Worth exploring when: you run a blog, a store or a micro SaaS and want AI assistants to find your newest content right away.
Start small: one read-only tool on the WordPress REST API → test it as a custom connector → add search and latest posts → publish a docs page.
Do not skip: read-only labels, no secret keys in public tools, a clear privacy policy, and testing every tool before you share the URL.
Final Thought
Search engines are one door. AI assistants are the new one. An MCP keeps that door open the moment something new is published.
Try the Worksmarto MCP, and get the next experiment first in the Worksmarto newsletter.
Build it with AI. Run it on free APIs. Let AI assistants find you first.