TCXC MCP Server
Enables AI assistants to interact with the TelecomsXChange wholesale telecom API for voice termination, SMS, DID management, eSIM, and account operations through 55 tools and 7 guided workflows using natural language.
README
TCXC MCP Server
The TCXC MCP Server is an open-source Model Context Protocol (MCP) server that connects AI assistants — Claude, OpenAI, Cursor, and any MCP-compatible client — to the TelecomsXChange (TCXC) wholesale telecommunications marketplace. It exposes 55 tools and 7 guided workflows for voice termination, SMS messaging, DID phone numbers, eSIM data packages, HLR lookups, and carrier account management, so telecom operators and developers can run wholesale telecom operations in plain English.
What Can AI Assistants Do With It?
| Capability | What your AI assistant can do |
|---|---|
| Voice Termination | Search carrier rates by prefix, run A/B route test calls, analyze CDRs and call quality (ASR/ACD) |
| SMS Services | Compare global SMS termination rates, review delivery history and traffic summaries |
| DID Numbers | Search, purchase, and configure virtual phone numbers (voice, SMS, fax capable) |
| eSIM Exchange | Browse data packages by country, purchase eSIMs, top up, and track orders |
| Number Intelligence | Live HLR lookups, portability data, and fraud/reputation scoring |
| Financial | Check buyer/seller balances and payment history in real time |
| Account Management | Provision SIP/SMS sub-accounts, manage IPs, block/unblock accounts |
| Analytics | Live calls, top routes, traffic summaries, and performance insights |
Example requests once connected:
"What's my TCXC account balance?"
"Search voice rates for prefix 44 and compare the top 3 carriers"
"Find eSIM data packages for Turkey and buy the cheapest 5GB option"
Quick Start
Three steps: install → add your TCXC credentials → connect your AI client.
1. Install
Requires Python 3.10+ and a TelecomsXChange (TCXC) or NeuTRAFIX (NTX) account.
# Clone and navigate
git clone https://github.com/TelecomsXChangeAPi/tcxc-mcp-server.git
cd tcxc-mcp-server
# Setup virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Create your config from the template
cp .env.example .env
2. Add Your TCXC Credentials
Get your API credentials from your dashboard:
- TCXC: members.telecomsxchange.com → Account → API Settings
- NeuTRAFIX: members.neutrafix.net → Account → API Settings
Edit .env:
# Buyer credentials (required)
TCXC_USERNAME=your_username
TCXC_API_KEY=your_api_key
# Seller credentials (required)
TCXC_SELLER_USERNAME=your_seller_username
TCXC_SELLER_API_KEY=your_seller_api_key
# API endpoint
# NeuTRAFIX users: change the base URL to https://apiv2.neutrafix.net
TCXC_API_BASE_URL=https://apiv2.telecomsxchange.com
Both buyer and seller credentials must be provided for the server to start.
3. Connect Your AI Client
The server speaks standard MCP over stdio and works with every major MCP client:
| AI Client | Setup |
|---|---|
| Claude Desktop | JSON config (below) |
| Claude Code | claude mcp add one-liner |
| OpenAI Codex CLI | codex mcp add one-liner |
| Cursor | JSON config (below) |
| Any MCP client | Point it at start_mcp_server.sh |
In every case, use absolute paths and restart the client after changing its config.
Claude Desktop
Add to the config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%/Claude/claude_desktop_config.json
{
"mcpServers": {
"tcxc": {
"command": "/absolute/path/to/tcxc-mcp-server/start_mcp_server.sh"
}
}
}
Claude Code (CLI)
claude mcp add tcxc /absolute/path/to/tcxc-mcp-server/start_mcp_server.sh
OpenAI Codex CLI
codex mcp add tcxc /absolute/path/to/tcxc-mcp-server/start_mcp_server.sh
Or add to ~/.codex/config.toml:
[mcp_servers.tcxc]
command = "/absolute/path/to/tcxc-mcp-server/start_mcp_server.sh"
Cursor
Add to ~/.cursor/mcp.json (or .cursor/mcp.json in your project):
{
"mcpServers": {
"tcxc": {
"command": "/absolute/path/to/tcxc-mcp-server/start_mcp_server.sh"
}
}
}
Any Other MCP Client
Point your client at the wrapper script (which loads credentials from .env):
command: /absolute/path/to/tcxc-mcp-server/start_mcp_server.sh
Or run the Python entry point directly and pass credentials as environment variables:
command: /absolute/path/to/tcxc-mcp-server/venv/bin/python
args: ["/absolute/path/to/tcxc-mcp-server/src/tcxc_mcp/server.py"]
env: TCXC_USERNAME, TCXC_API_KEY, TCXC_SELLER_USERNAME, TCXC_SELLER_API_KEY
Running Standalone
# Activate environment
source venv/bin/activate
# Start server
./start_mcp_server.sh
# Or run directly
python src/tcxc_mcp/server.py
Testing
Use the interactive client to test tools:
source venv/bin/activate
python src/tcxc_mcp/mcp_client.py
Interactive commands:
list_tools- Show all available toolscall_tool <name> <json_params>- Execute a toollist_prompts- Show guided workflowsget_prompt <name> <json_args>- Get workflow guidance
Available Tools
Voice Operations
search_voice_market(prefix="1", route_type="CLI")
get_voice_destinations(country="United States")
list_call_records(date_from="2026-01-01", date_to="2026-01-31")
test_voice_route(i_account="123", cld1="+12125551234", cli1="+12025551234", i_connection1="8721", ...)
SMS Operations
marketview_search_sms(prefix="44", country="United Kingdom")
buyers_smshistory(date_from="2026-01-01", date_to="2026-01-31")
buyers_sms_summary(date_from="2026-01-01", date_to="2026-01-31")
DID Management
number_market(country="United States", did_type="mobile")
number_list(prefix="1")
number_purchase(i_did="12345", billing_i_account="123", contact="sip:user@domain.com")
eSIM Exchange
esim_countries()
esim_market(country="Turkey")
esim_purchase(i_esim_package="456", i_account="123")
esim_list()
esim_topup_list(iccid="8944500000000000000")
esim_topup_purchase(iccid="8944500000000000000", package_type_id="789", i_account="123")
esim_orders(date_from="2026-01-01", date_to="2026-01-31")
Financial & Account
buyers_balance()
seller_balance()
buyers_payhistory(date_from="2026-01-01", date_to="2026-01-31")
buyers_account_list()
Utilities
hlr_lookup(number="+12125551234", i_account="123")
get_number_score(phone_number="+12125551234")
tools_country_codes()
Full tool list: 55 tools available. See tool documentation for details.
Guided Workflows (Prompts)
The server includes intelligent prompts for complex tasks:
- search_best_voice_rates - Find and compare voice carriers
- monitor_account_health - Comprehensive account health check
- purchase_did_workflow - Guided DID purchase process
- analyze_call_quality - CDR analysis and quality insights
- sms_market_research - SMS carrier comparison
- validate_routing_setup - Test and validate phone routing
- compare_carriers - Head-to-head carrier comparison
Use prompts in Claude: "Use the monitor_account_health prompt to check my TCXC account"
See PROMPTS.md for the full guide.
Troubleshooting
Server won't start
# Check environment variables
cat .env
Confirm all four credential variables are set (buyer and seller).
Authentication fails
# Test credentials directly
curl -u "$TCXC_USERNAME:$TCXC_API_KEY" \
--digest \
"$TCXC_API_BASE_URL/buyers/balance"
Import errors
# Reinstall dependencies
pip install -r requirements.txt
# For HTTP/2 support
pip install "httpx[http2]"
Claude Desktop connection issues
- Use absolute paths (no
~or relative paths) - Restart Claude Desktop after config changes
- Check Claude logs:
tail -f ~/.config/claude/mcp.log
Development
# Install dev dependencies
pip install -e ".[dev]"
# Format code
make format
# Run linting
make lint
# Type checking
make typecheck
# Run tests
make test
Architecture
- Framework: FastMCP (official MCP Python SDK)
- HTTP Client: HTTPX with HTTP/2 and digest auth
- Authentication: Dual account support (buyer/seller)
- Security: Input validation, TLS 1.2+, rate limiting
- Performance: Async/await, connection pooling
Security
- Credentials stored in
.env(never committed to git) - File permissions:
chmod 600 .envandchmod 700 start_mcp_server.sh - All inputs validated and sanitized before API calls
- TLS verification enforced (TLS 1.2+)
- Error messages sanitized to prevent information disclosure
See SECURITY.md for detailed security practices.
Environment Support
Production: https://apiv2.telecomsxchange.com
- Real billing and provisioning
- Use production credentials
Sandbox: Contact TCXC support for sandbox endpoint
- Safe testing environment
- No real billing impact
Frequently Asked Questions
What is the TCXC MCP Server?
The TCXC MCP Server is an open-source Python implementation of the Model Context Protocol that gives AI assistants direct, secure access to the TelecomsXChange wholesale telecom API. Instead of writing REST integration code, you connect the server to Claude, OpenAI, or another MCP client and ask for what you need in natural language — search carrier rates, test voice routes, buy DID numbers or eSIMs, run HLR lookups, and monitor traffic. It ships with 55 tools and 7 guided workflow prompts, uses HTTP Digest Authentication with TLS 1.2+, and validates every input before it reaches the API.
Which AI assistants and tools work with it?
Any MCP-compatible client. Tested configurations include Claude Desktop, Claude Code (CLI), OpenAI Codex CLI, and Cursor. The server communicates over standard MCP stdio, so agent frameworks that support MCP — such as the OpenAI Agents SDK and Claude Agent SDK — can use it too.
What do I need to get started?
Three things: Python 3.10 or newer, a TelecomsXChange or NeuTRAFIX account, and your API credentials from the member dashboard (Account → API Settings). Setup takes about five minutes: clone the repo, install dependencies, copy .env.example to .env, add your credentials, and register the server with your AI client.
Does it work with NeuTRAFIX?
Yes. NeuTRAFIX runs the same platform API. Set TCXC_API_BASE_URL=https://apiv2.neutrafix.net in your .env and use your NTX credentials — everything else works identically.
Is it free to use?
The server itself is free and MIT-licensed. API usage is billed by TelecomsXChange at standard marketplace rates — voice, SMS, DID, and eSIM purchases made through the tools are real transactions on your TCXC account.
Can an AI agent make purchases autonomously?
Purchase tools (number_purchase, esim_purchase, esim_topup_purchase) execute real transactions against your account balance. MCP clients like Claude Desktop ask for your approval before each tool call by default; we recommend keeping approval prompts enabled for purchase and account-management tools.
How is it secured?
Credentials stay local in a gitignored .env file and are sent only to the TCXC API over HTTPS using HTTP Digest Authentication. The client enforces TLS 1.2+ with strong cipher suites, whitelists API endpoints, validates and sanitizes all inputs, rate-limits requests, and strips sensitive data from error messages and logs. See SECURITY.md.
What is TelecomsXChange?
TelecomsXChange is a wholesale telecommunications marketplace where licensed carriers, operators, and communication service providers buy and sell voice termination, SMS, DID numbers, and eSIM data packages. Buyers get transparent per-route pricing with quality metrics (ASR/ACD); sellers reach a global base of interconnected buyers through one API.
License
MIT License - see LICENSE file
Support
- Issues: GitHub Issues
- TCXC Support: telecomsxchange.com/contact
- API Docs: apidocs.telecomsxchange.com
- Dashboard: members.telecomsxchange.com
Contributing
Contributions welcome! Please:
- Fork the repository
- Create a feature branch
- Make your changes with tests
- Submit a pull request
Built with the MCP Python SDK | Powered by TelecomsXChange
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。