@saperly/mcp
MCP server for Saperly that enables AI agents to provision phone numbers, place calls, send SMS, and manage credentials via 36 tools backed by the Saperly API.
README
@saperly/mcp
Model Context Protocol server for Saperly. Gives any MCP-compatible AI agent (Claude Desktop, Claude Code, Cursor) the ability to provision phone numbers, place calls, send SMS, and manage credentials.
32 tools across 12 groups, all backed by the public Saperly API.
Install
npm install -g @saperly/mcp
Or run on demand with npx @saperly/mcp.
Quickstart
Register the server with Claude Desktop. Add this block to claude_desktop_config.json:
{
"mcpServers": {
"saperly": {
"command": "npx",
"args": ["-y", "@saperly/mcp"],
"env": { "SAPERLY_API_KEY": "sk_live_..." }
}
}
}
For Claude Code:
claude mcp add saperly -- npx -y @saperly/mcp
export SAPERLY_API_KEY=sk_live_...
Get an API key at saperly.com/settings/keys. For key-management tools (saperly_key_*) set SAPERLY_API_KEY to a service key instead.
Use as a hosted connector
Same auth model, no local install — the hosted connector serves the OAuth-usable surface (the service-key saperly_key_* tools stay local-only). The hosted endpoint at https://saperly.com/api/v1/mcp speaks the streamable HTTP transport and accepts three credential types: API-key bearer, opaque connector URL, and OAuth 2.1 access token (Saperly is the issuer).
- Anthropic Messages API (
mcp_servers): point athttps://saperly.com/api/v1/mcpand pass your API key asauthorization_token. - Claude.ai custom connector — OAuth tab: paste
https://saperly.com/api/v1/mcp. Claude does the discovery + dynamic registration + PKCE dance and lands the user on a consent screen at saperly.com. - Claude.ai custom connector — No auth tab: mint a connector URL at saperly.com/settings/connectors and paste it with auth set to None.
- ChatGPT — custom connector: add
https://saperly.com/api/v1/mcp; ChatGPT runs the same OAuth discovery + registration flow as Claude.ai. - CrewAI:
pip install "saperly[crewai]"then wrap withSaperlyTools(api_key=…). See the Python SDK.
Tool groups
The canonical tool list lives in manifest.json. One sentence per group:
- lines: Provision and manage phone numbers (modes: webhook, audio, hosted).
- calls: Start and control calls, wait for inbound calls, retrieve history, run hosted AI conversation calls.
- messages: Send outbound SMS and read inbound conversation history.
- keys: Mint, rotate, and revoke child API keys (service-key auth required).
- consent: Grant, check, and revoke TCPA outbound consent per phone number.
- disclosures: Configure TCPA disclosure scripts played at call start.
- billing: Read account balance and transactions.
- account: Whole-account snapshot covering lines, balance, usage, and recent calls.
- webhooks: List delivery attempts, read aggregate stats, send test webhooks.
- usage: Aggregate call minutes, SMS counts, and costs by day or month.
- settings: Read or update account-level configuration (e.g. default webhook URL).
- voices: List TTS voices available for hosted-mode calls.
Auth
Saperly uses a two-tier credential model. Both authenticate via SAPERLY_API_KEY on the MCP server's environment.
| Credential | Use case |
|---|---|
API key (sk_live_…, sk_test_…) |
Place calls, send SMS, manage lines, read usage. |
Service key (sk_svc_live_…, sk_svc_test_…) |
Mint, rotate, and revoke child API keys via the saperly_key_* tools. Credential management only. |
Full reference at docs.saperly.com/service-keys.
Resources
- Docs: https://docs.saperly.com
- Quickstart: https://docs.saperly.com/quickstart
- Agent onboarding: https://docs.saperly.com/agent-onboarding
- Service keys: https://docs.saperly.com/service-keys
- Patterns and pitfalls: ./AGENTS.md
- API reference: https://docs.saperly.com/api-reference
- Issues: https://github.com/Saperly/saperly-mcp/issues
- Discord: invite link in our docs
License
MIT. See LICENSE.
Contact
hello@saperly.com or open an issue.
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。