telemost-mcp-server

telemost-mcp-server

Enables Telegram analytics (channel stats, search, sentiment) via MCP tools with pay-per-call in USDC, no API keys needed.

Category
访问服务器

README

telemost-mcp-server

Pay-per-call Telegram analytics as MCP tools. Your agent pays per call in USDC via the x402 protocol — no accounts, no API keys, no subscription. Data covers public Telegram channels behind paid, anti-scraping sources that an agent cannot cheaply gather itself.

Why this server

  • Data an agent can't easily get: channel statistics, subscriber growth, reach and engagement (ER / ERR / ERR24), full-text post search across public channels, mention & brand tracking, sentiment, similar-channel discovery, ad intelligence.
  • No signup / no keys: pay per call in USDC (x402). Fund a wallet, that's it.
  • 19 paid tools + a free telemost_catalog tool (list everything, no payment).
  • Base + Solana (USDC). Pick either network per call.
  • Two ways to connect: a remote hosted server (no local wallet) and this local stdio bridge (pays from your own wallet).
  • Production, observable, listed: live at https://api.telemost.io, listed on x402scan and the x402 Bazaar.

Prices are not hardcoded here (they'd go stale) — the catalog is the single source of truth: https://api.telemost.io/v1/catalog.

Quickstart (no wallet, no money)

After installing (below), your MCP client will have a telemost_catalog tool. Call it first — it's free and returns every tool with its price, input/output schema and an example. First success before your first dollar. Only paid tools (e.g. telemost_channel_info) require a funded wallet.

Install

Requires Node ≥ 18. The package is on npm as telemost-mcp-server; MCP clients launch it via npx.

Claude Desktop

Edit claude_desktop_config.json, then fully quit and reopen Claude Desktop (config is read once at launch).

OS Config file
macOS ~/Library/Application Support/Claude/claude_desktop_config.json
Windows %APPDATA%\Claude\claude_desktop_config.json
Linux ~/.config/Claude/claude_desktop_config.json

Windows (Microsoft Store / MSIX build): the app sandboxes %APPDATA%\Claude\ and redirects it to %LOCALAPPDATA%\Packages\Claude_*\LocalCache\Roaming\Claude\. Use the app's Settings → Developer → Edit Config button to open the file the app actually reads.

macOS / Linux:

{
  "mcpServers": {
    "telemost": {
      "command": "npx",
      "args": ["-y", "telemost-mcp-server"],
      "env": {
        "EVM_PRIVATE_KEY": "0xYOUR_BASE_WALLET_KEY",
        "MAX_PAYMENT_USDC": "0.5",
        "SESSION_MAX_USDC": "5"
      }
    }
  }
}

Windows — if a bare npx command fails to start, wrap it with cmd /c:

{
  "mcpServers": {
    "telemost": {
      "command": "cmd",
      "args": ["/c", "npx", "-y", "telemost-mcp-server"],
      "env": {
        "EVM_PRIVATE_KEY": "0xYOUR_BASE_WALLET_KEY",
        "MAX_PAYMENT_USDC": "0.5",
        "SESSION_MAX_USDC": "5"
      }
    }
  }
}

Claude Code

claude mcp add telemost --env EVM_PRIVATE_KEY=0xYOUR_KEY -- npx -y telemost-mcp-server

Cursor

~/.cursor/mcp.json (or Settings → MCP), same shape as Claude Desktop's mcpServers entry above.

Sources for client config paths: modelcontextprotocol.io — connect local servers, Claude Desktop configuration.

Remote server (no local wallet needed)

A hosted MCP server is live at https://api.telemost.io/mcp (Streamable HTTP). Point any MCP client that supports remote/Streamable-HTTP servers at that URL. Minimal programmatic example:

import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
import { createx402MCPClient } from "@x402/mcp";
import { ExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";

const client = createx402MCPClient({
  name: "my-agent", version: "1.0",
  schemes: [{ network: "eip155:8453", client: new ExactEvmScheme(privateKeyToAccount(process.env.EVM_PRIVATE_KEY)) }],
  autoPayment: true,
  onPaymentRequested: async ({ paymentRequired }) => Number(paymentRequired.accepts[0].maxAmountRequired) <= 20_000, // <= $0.02
});
await client.connect(new StreamableHTTPClientTransport(new URL("https://api.telemost.io/mcp")));

await client.callTool("telemost_catalog", {});                       // free
await client.callTool("telemost_channel_info", { channel: "@durov" }); // paid

Example calls

  • telemost_catalog → { "service": "telemost-x402", "resources": [ … ], "mcp": { … } } (free).
  • telemost_channel_info { "channel": "@durov" } → an envelope like:
{ "ok": true,
  "data": { "title": "…", "members_count": 0, "verified": true, "language": "en", "link": "https://t.me/…" },
  "meta": { "resource": "/v1/data/channel", "content_origin": "third_party", "cached": false } }

Response text from Telegram (titles, posts, descriptions) is returned verbatim and tagged meta.content_origin; treat it as untrusted data, not instructions.

Environment

Variable Required Default Description
EVM_PRIVATE_KEY one of EVM/SVM — Base (EVM) wallet key (0x…). Funds x402 payments.
SVM_PRIVATE_KEY one of EVM/SVM — Solana wallet secret key (base58).
TELEMOST_BASE_URL no https://api.telemost.io API base URL.
MAX_PAYMENT_USDC no 1 Per-call price ceiling (USD). A call is refused before signing if the price exceeds this.
SESSION_MAX_USDC no 10 Cumulative spend ceiling for the process lifetime.
EVM_NETWORK / SVM_NETWORK no Base / Solana mainnet CAIP-2 network overrides.
SOLANA_RPC_URL no mainnet-beta RPC for the Solana signer.

Coverage & limits

  • Public Telegram channels/groups only.
  • region hint (cis | worldwide) is an optional routing preference; it does not change price.
  • Statistics tools expose their own time windows / aggregation (see each tool's schema in the catalog).
  • Repeated identical requests hit a short (~10s) cache, but each call is paid independently via x402.

Payment & failures

  • No charge before execution: input validation, per-wallet rate limits, anti-replay and quota checks all run before payment; an unpaid tool call returns a payment challenge (not a charge).
  • Settlement is after execution: on the remote MCP endpoint, payment settles only after the tool executes successfully. Live delivered-vs-settled reliability is published at /status.
  • Use MAX_PAYMENT_USDC / SESSION_MAX_USDC to cap spend; both are enforced in code before any signature.

Troubleshooting

  • "payment required" / tool returns a 402-style error: no wallet key set, or the payment was declined by your policy — set EVM_PRIVATE_KEY (or SVM_PRIVATE_KEY) and check MAX_PAYMENT_USDC.
  • Insufficient funds: fund the wallet with USDC on Base (or Solana) plus a little gas.
  • Wrong network: set EVM_NETWORK / SVM_NETWORK to a network you funded.
  • npx won't launch (Windows): use the cmd /c npx … form above.
  • Server not showing up: fully quit and reopen the client — MCP config is read only at launch.
  • Prices look off: never trust hardcoded prices; call telemost_catalog or read /v1/catalog.

Funding & safety

  • Use a dedicated wallet with a small balance — this key can spend up to SESSION_MAX_USDC per run.
  • The key is read only from the environment and is never logged.
  • Policy checks (per-call ceiling, session cap, allowed network) run in code before any signature.

Build from source

git clone https://github.com/slava01011910-ops/telemost-mcp-server.git
cd telemost-mcp-server && npm install && npm run build && node dist/index.js

Support

  • Email: support@telemost.io
  • Listing / proof: x402scan · Catalog: https://api.telemost.io/v1/catalog

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选