MCP AI Agent: Build a Model Context Protocol Agent on Telegram (Ready Prompt)

A "MCP Hello Page" just hit Hacker News — and it's the simplest proof that Model Context Protocol is the missing link between AI agents and real-world tools. Here's how to build an agent that uses MCP-style tool calling on Telegram, with a ready-to-use prompt.

What Is MCP (Model Context Protocol)?

Model Context Protocol is an open standard — originally built by Anthropic — that lets AI agents talk to external tools, APIs, databases, and file systems through a unified interface. Think of it as USB-C for AI tools: one protocol that connects any agent to any resource.

The "MCP Hello Page" by Hybrid Logic (currently trending on Hacker News with 50+ points) demonstrates just how simple an MCP integration can be. It's a single web page that serves as a minimal MCP server — proving that the barrier to entry is practically zero.

Why does this matter? Because up until now, every AI agent that wanted to do something useful — search the web, check a database, call an API, read a file — had to reinvent the wheel. Custom integrations, hardcoded endpoints, fragile prompt engineering. MCP standardizes all of it. Your agent speaks MCP once, and it can talk to any MCP-compatible server.

Why a Telegram MCP Agent Changes the Game

Everyone's talking about MCP at the protocol level — servers, clients, transports. But the most practical use case is staring us in the face: a Telegram bot that acts as an MCP client.

Think about what that unlocks:

• Search the web — "Find the latest Rust blog posts about async" → agent calls a search tool → returns results
• Query your data — "How many open issues in our repo?" → agent calls GitHub API → returns real numbers
• Run reports — "Summarize this week's HN top stories" → agent calls content fetcher → returns summary
• Take action — "Send a follow-up to the Acme deal" → agent calls email tool → confirms sent

No custom code. No webhooks to configure. No brittle API wrappers. Just one prompt that tells the agent what tools it has, and the agent figures out when and how to use them.

The Ready-to-Use MCP Agent Prompt

Copy this prompt into your OpenClaw-based Telegram AI agent. It gives the agent a full MCP-style tool-use capability — with built-in resources, tools, and a structured reasoning process.

You are an MCP (Model Context Protocol) agent running on Telegram.

You have access to the following tools and resources:

## Available Tools

### Tool: web_search
- Description: Search the web for current information
- Usage: Call when you need up-to-date facts, news, documentation, or anything not in your training data
- Expected input: search query string
- Expected output: search results with titles, URLs, and snippets

### Tool: web_content
- Description: Fetch and extract readable content from a URL
- Usage: Call to read articles, documentation pages, blog posts
- Expected input: URL
- Expected output: markdown-formatted page content

### Tool: code_evaluate
- Description: Run evaluation on code, data, or logic
- Usage: Call to validate outputs, check reasoning, or test hypotheses
- Expected input: the code or logic to evaluate
- Expected output: evaluation result

## Decision Protocol

When the user sends a request:

1. **Scope** — Identify what the user actually needs. Is it information? Analysis? Action?
2. **Plan** — Decide which tool(s) to use and in what order
3. **Execute** — Call the first tool and examine output
4. **Synthesize** — Combine tool results into a coherent response
5. **Verify** — Check if the answer fully addresses the user's request. If not, iterate.

## Output Rules

- Always cite sources when using web_search or web_content
- If a tool returns an error, explain what happened and offer alternatives
- When comparing options (products, frameworks, services), present a structured table
- For Telegram delivery: keep responses scannable — use bullet lists, short paragraphs, and bold highlights
- Never fabricate tool results. If you can't call a tool, say so honestly.

## Tool-Use Examples

### Example 1: Research task
User: "What's new in the Rust ecosystem this week?"
You: "Let me search for that." [call web_search] → present findings with sources

### Example 2: Content summarization
User: "Summarize this article: https://example.com/post"
You: "Let me fetch the article." [call web_content] → provide structured summary

### Example 3: Comparative analysis
User: "Which AI coding assistant is best for TypeScript?"
You: "Let me gather current data." [call web_search] → compare with table → recommend

---

Start each response with an appropriate tool call if the user's request requires up-to-date information or external data. For simple Q&A within your knowledge, answer directly.
        

How to Use It

1. Deploy on GetClawCloud — paste the prompt into your OpenClaw Agent Skills and set the system prompt
2. Paste the prompt — copy the prompt above and set it as your agent's SKILL.md
3. Send to test — message your Telegram bot with "Search for the latest AI agent frameworks" and watch it use its tools

Why MCP Matters for AI Agents on Telegram

The MCP Hello Page on Hacker News proves something important: the barrier to entry for MCP is remarkably low. A single-page HTML file can serve as an MCP server. That means any developer with basic web skills can build tools that AI agents can discover and use.

For Telegram AI agents specifically, MCP unlocks the most requested features:

• Real-time data — your bot isn't limited to its training cutoff
• Action execution — your bot can actually do things, not just talk about them
• Composability — plug in new tools without retraining or rewriting prompts
• Interoperability — the same agent works with any MCP-compatible server, anywhere

The future of AI agents isn't bigger models. It's better connections. MCP is how you give your agent hands — not just a mouth.

MCP vs. Function Calling: What's the Difference?

If you've been following AI agent development, you've seen "function calling" in GPT-4, Claude, and Gemini. MCP is different in a few key ways:

For a Telegram agent, MCP's biggest advantage is dynamic discovery. You don't need to enumerate every tool in your prompt upfront. The agent asks the MCP server "what do you offer?" and gets a complete list. Add a new server tomorrow? The agent finds it automatically.

Real Use Cases for Your MCP Telegram Agent

Once your Telegram agent has MCP-style tool access, here's what it can actually do:

Daily News Briefing

"Give me a morning briefing of tech news." The agent searches top tech news sites, fetches articles, summarizes key stories, and sends a digest — all from a single message.

Competitor Research

"What's new with our top 3 competitors this week?" The agent searches for each competitor, fetches relevant pages, and produces a structured comparison report.

DevOps Monitoring

"Check if there are any major outages right now." The agent searches status pages and news sources, then reports current incidents.

Learning Assistant

"I want to learn Rust for web assembly — find the best learning path." The agent searches for resources, reads course descriptions, and recommends a structured learning plan with links.

Take This Further

The beauty of MCP is that it's an extensible protocol. You can start with web search and content fetching (as in the prompt above), then add:

• A database MCP server — query your Postgres or SQLite database from Telegram
• A filesystem MCP server — read/write files in a sandboxed directory
• A GitHub MCP server — create issues, review PRs, check CI status
• An email MCP server — send and search emails through your agent

Each new server is a new capability your agent unlocks — no prompt changes required.