Lokahi MCP Server
Enables AI assistants to search for wellness practitioners, book sessions, manage offerings, and more via the Lokahi API, with OAuth and API key authentication.
README
Lokahi MCP Server
Remote MCP server for the Lokahi wellness marketplace API, deployed as a Cloudflare Worker.
Lets AI assistants (Claude, ChatGPT, Cursor, etc.) search for practitioners, book sessions, manage offerings, and more — on behalf of authenticated users.
Setup
npm install
KV Namespace
Create the KV namespace for OAuth state storage:
wrangler kv namespace create OAUTH_KV
Copy the output ID and update wrangler.jsonc — replace PLACEHOLDER_KV_ID with the actual namespace ID.
Secrets
Set the shared secret used for the Worker-to-Anvil private handshake (must match the value configured in Anvil):
wrangler secret put ANVIL_OAUTH_SECRET
Development
npm run dev
Deploy
npm run deploy
Authentication
The MCP server supports two authentication methods:
OAuth (recommended for Claude.ai / ChatGPT)
Select OAuth in your connector settings. The Worker handles the full OAuth 2.0 flow automatically:
- You are redirected to Lokahi to log in and grant consent.
- After approval, the Worker exchanges the authorization for tokens.
- Access tokens expire after 1 hour; refresh tokens last 30 days and re-validate with Anvil that the user is still active.
The Worker implements spec-compliant OAuth with PKCE, discovery (/.well-known/oauth-authorization-server), and dynamic client registration (/oauth/register).
API Key (direct — for Claude Desktop, Cursor, Claude Code)
Generate an API key from your Lokahi account at Settings > API Access on app.lokahi.life. Keys start with lok_live_ and are shown once at creation time.
The server accepts the API key in two ways (checked in order):
Authorization: Bearer lok_live_...header (preferred)?apiKey=lok_live_...query parameter (fallback)
Public tools (search, practitioner profiles) work without auth. Seeker and practitioner tools require authentication.
Connecting
Claude.ai (remote MCP)
- Go to Claude.ai > Settings > Connectors > Add MCP Server
- URL:
https://lokahi-mcp.<your-subdomain>.workers.dev/sse - Authentication: select OAuth — the login flow is handled automatically
Claude Desktop / Cursor / Claude Code
Add to your MCP config (e.g. ~/.claude/claude_desktop_config.json):
{
"mcpServers": {
"lokahi": {
"url": "https://lokahi-mcp.<your-subdomain>.workers.dev/sse",
"headers": {
"Authorization": "Bearer lok_live_YOUR_KEY_HERE"
}
}
}
}
ChatGPT (GPT Actions / Connectors)
ChatGPT supports OAuth connectors. Point it at the Worker URL and select OAuth authentication — the discovery endpoint at /.well-known/oauth-authorization-server provides all the configuration automatically.
Alternatively, use the OpenAPI spec directly:
https://app.lokahi.life/_/api/v1/openapi.json
Available Tools
Public (no auth)
search_practitioners— Search by keyword, specialty, or locationget_practitioner— Get full profile by slugget_available_dates— Dates with availability in a monthget_available_slots— Time slots on a specific datelist_modalities— All wellness modalitieslist_specialties— All specialties
Seeker (authenticated)
list_bookings— Your bookingsget_booking— Booking detailscreate_booking— Book a sessioncancel_booking— Cancel a bookingreschedule_booking— Reschedule a bookinglist_favourites— Your favourite practitionerstoggle_favourite— Add/remove favouritelist_conversations— Your conversationsget_messages— Messages in a conversationsend_message— Message a practitioner
Practitioner (authenticated + practitioner role)
practitioner_dashboard— Dashboard overviewlist_offerings/create_offering/update_offering/delete_offeringlist_practitioner_bookings/get_practitioner_bookingaccept_booking/decline_booking/practitioner_cancel_booking/complete_bookingget_availability/update_availabilityadd_availability_override/delete_availability_overridelist_practitioner_conversations/practitioner_send_messageget_practitioner_profile/update_practitioner_profile
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。