crowntowncompost-mcp
MCP server for the Crown Town Compost customer portal that lets you check pickups, invoices, and upcoming collections, skip services, report missed pickups, and update account details through natural language.
README
crowntowncompost-mcp
MCP server for the Crown Town Compost customer portal — check your pickups, invoices, and upcoming collection days, skip a service, report a missed pickup, and update your account, all in natural language.
Developed and maintained by AI (Claude Code). Use at your own discretion.
What it talks to
crowntowncompost.com is a marketing site with no customer data. Everything lives at
portal.crowntowncompost.com, a Django app (a white-labeled StopSuite hauler platform).
This server signs in server-side with your own portal username and password — a normal form
login that returns a session cookie — and reads the same pages and JSON endpoints the website uses.
No browser extension, no bot-wall workaround, no third-party service in the middle.
Install
npm install -g crowntowncompost-mcp
Then add it to your MCP host with your portal credentials:
{
"mcpServers": {
"crowntowncompost": {
"command": "npx",
"args": ["-y", "crowntowncompost-mcp"],
"env": {
"CROWNTOWN_USERNAME": "you@example.com",
"CROWNTOWN_PASSWORD": "your-portal-password"
}
}
}
}
Locally you can instead copy .env.example to .env. The server boots without credentials (so a
host's install-time probe succeeds); the configuration error surfaces on the first tool call.
Tools
Reads
| Tool | What it returns |
|---|---|
crowntown_healthcheck |
Whether credentials work, plus account status — distinguishes "no creds" from "bad creds" from "site error" |
crowntown_get_dashboard |
Account status, subscription (plan, price, renewal date), next service date, service addresses + pickup days, and your environmental impact |
crowntown_get_account |
Contact details and notification preferences |
crowntown_get_pickup_schedule |
Pickup day(s) and time window per address — the official set-out-by time plus an observed arrival window (earliest/latest/typical, consistent vs varies) derived from your collection history |
crowntown_list_service_history |
Past collection stops — date, outcome, time, weight, services. Paginated; filter by success/missing/empty/inaccessible/unacceptable |
crowntown_list_upcoming_services |
Upcoming collection days, each with the ids needed to skip it |
crowntown_list_invoices |
Billing history with amounts, status, and Stripe payment links |
Writes (all confirm-gated)
| Tool | What it does |
|---|---|
crowntown_skip_service |
Skip or un-skip an upcoming collection day |
crowntown_update_account |
Update contact details / notification preferences |
crowntown_report_missed_pickup |
Report that a collection was missed |
crowntown_contact_support |
Send a message to customer support |
Every mutating tool takes confirm. Without confirm: true it makes no network call and returns
a dry-run preview of exactly what would be sent. Where a re-read can prove the change stuck (skips,
account updates) the tool re-reads and reports verified; where it can't (support messages, missed-pickup
reports) it says so rather than claiming success.
Payments are deliberately out of scope — crowntown_list_invoices returns the hosted invoice URL for
you to open in a browser.
Hosted connector (claude.ai)
There is also a Cloudflare Worker entry point (src/worker.ts) that makes this server reachable from
claude.ai — web, desktop, and mobile — rather than only from a local stdio host. It wraps the same
tools behind an OAuth login page where you enter your own portal credentials.
This works because the portal's auth is a plain server-side form login, with nothing that needs a browser or a filesystem. Since the portal issues no refresh token, your credentials are stored encrypted so the connector can sign in again when the session expires.
Deploy is manual — see docs/DEPLOY-CONNECTOR.md.
Without the MCP
The skills/crowntown-portal skill does the same things with curl in a shell — useful in scripts or
on a machine where this server isn't installed. It documents the Django CSRF handshake, both JSON
endpoints, and every write's field list.
Development
npm install
npm run build
npm test
Tests mock the network — no credentials needed and nothing hits the live portal. tests/server-boot.test.ts
spawns the real built artifacts (the npm bin and the bundle, the latter without node_modules) and runs
the MCP handshake against them.
npm run worker:test runs the connector suite separately, inside the real Workers runtime — npm test
excludes it, since it imports cloudflare:test and can't load under Node.
Endpoint shapes are documented in docs/CROWNTOWN-API.md, including which
parts are live-verified and which are not.
License
MIT
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。