BlockBeak: Building Your Own ‘Perplexity for Crypto’ Telegram BotBlockBeak: Building Your Own ‘Perplexity for Crypto’ Telegram Bot

01.TL;DR

BlockBeak is a prototype Telegram bot demonstrating how simple yet powerful AI applications can be built with the formula:

LLM + Prompt + Heurist Mesh tools = Functional AI Agent

The entire project runs on just ~200 lines of Python code, leveraging the OpenAI Agents SDK and Heurist Mesh’s Model Context Protocol (MCP). Try it in the Heurist Ecosystem Builder Telegram group or fork our open-source repo to build your own.

02.The Core Concept

Most AI implementations involve complex architectures and extensive engineering. BlockBeak proves that with the right building blocks—an LLM, a clear system prompt, and MCP-enabled agents—you can deliver a full-featured crypto research bot with minimal code.

03.How BlockBeak Works

The core logic fits in about 200 lines of Python:

1. Establish MCP Connection

   self.mcp_server = MCPServerSse(
       name="MCP SSE Server",
       params={"url": self.mcp_sse_url},
       client_session_timeout_seconds=60
   )
   

2. Instantiate the Agent

   agent = OpenAIAgent(
       name="BlockBeak Agent",
       instructions=self.instructions,
       mcp_servers=[self.mcp_server],
       model=self.model,
       model_settings=model_settings
   )
   

3. Process Messages

   async with self.mcp_server:
       result = await self._execute_with_retry(
           Runner.run,
           starting_agent=agent,
           input=message,
           context=self.context,
       )
   

Under the hood, retry logic and OOP patterns ensure reliability—but the essence remains refreshingly straightforward.

04.The Power of System Prompts

BlockBeak’s behavior is entirely driven by its system prompt, which defines:

• Agent capabilities and limits
• When and how to invoke each tool
• Desired response formats

Tweak the prompt and you can transform BlockBeak’s personality and scope without touching any Python code.

05.Heurist Mesh Agents in Action

BlockBeak taps into a diverse ecosystem of specialized agents:

• Bitquery Solana Token Info Agent
• CoinGecko Token Info Agent
• Cookie Project Info Agent
• DexScreener Token Info Agent
• Exa Search Agent
• Firecrawl Search Agent
• Funding Rate Agent
• MindAI KOL Agent
• Moni Twitter Insight Agent
• PumpFun Token Agent
• Solana Wallet Agent
• Truth Social Agent
• Twitter Info Agent
• Zerion Wallet Analysis Agent

Each brings domain-specific data—on-chain metrics, social sentiment, funding rates, portfolio analysis—to power comprehensive crypto insights.

06.Why This Matters

1. Development Simplicity: Build powerful AI tools with ~200 lines of code
2. Seamless Integration: Heurist Mesh works out-of-the-box with OpenAI Agents SDK
3. Tool Composability: Mix & match agents to cover any crypto research need
4. Prompt-Driven Customization: Change behavior via prompts, not code refactors

07.Try It Yourself

1. Clone the repo
2. Configure your MCP server at mcp.heurist.ai
3. Customize the system prompt for your use case
4. Deploy to your Telegram group

08.BlockBeak in Action

Let’s look at some examples of what BlockBeak can do when tapping into the Heurist Mesh agent ecosystem:

Example 1: Deep Research on EIPs

Ask “Do a deep dive on EIPs related to intent” and BlockBeak returns structured details on EIP-7521, ERC-4337, relevant links, and governance discussions.

Example 2: Token Analysis

Ask “Analyze HEU token” and receive cross-validated market data, liquidity insights, social media sentiment, and official references—all aggregated by multiple agents.

09.Moving Forward

BlockBeak is a proof-of-concept for a future where AI agents can be built and deployed rapidly with minimal engineering. By combining an LLM, a well-crafted prompt, and the right mesh of tools, developers can spin up domain-optimized assistants for any niche—crypto or beyond.

For more on the Model Context Protocol and agent orchestration, visit mcp.heurist.ai.

Fork it and start building: github.com/heurist/blockbeak

Try it live: t.me/HeuristEcosystemBuilder