avemguvern
Provides read-only access to the current government status of Romania via the get_government_status tool.
README
avemguvern.ro
A one-question site: does Romania currently have a (full, non-interim) government? Current answer: Nu! — Inca e interimar Bolojan!
Everything runs in a single Cloudflare Worker (free tier):
- Static page — minimalist verdict, served from
public/. - Public API —
GET /api/status(read),POST /api/status(admin, token-protected). - Crowd-sourced jokes —
POST /api/joke(public, body{"combo":["AUR","UDMR"],"joke":"..."}) lets visitors submit a joke for a party combination; stored but never shown live. Review them withGET /api/jokes(admin token):curl .../api/jokes -H "Authorization: Bearer $ADMIN_TOKEN". - MCP server — read-only
get_government_statustool at/mcp(Streamable HTTP).
State is one Cloudflare KV key (current). If KV is empty the Worker serves
DEFAULT_STATUS from src/index.ts, so it works before seeding.
Project layout
public/index.html # the page (no build step)
src/index.ts # Worker: routes /api/status, /mcp, else static assets
wrangler.jsonc # Worker + assets + KV config
seed.json # initial KV value
Local development
npm install
npx wrangler dev
Then:
# read
curl http://localhost:8787/api/status
# MCP: list tools
curl -X POST http://localhost:8787/mcp \
-H 'content-type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
# MCP: call the tool
curl -X POST http://localhost:8787/mcp \
-H 'content-type: application/json' \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_government_status"}}'
For a local admin token during wrangler dev, create .dev.vars:
ADMIN_TOKEN=some-local-token
Deploy
npm install
npx wrangler login
# 1. Create the KV namespace, then paste the returned id into wrangler.jsonc
npx wrangler kv namespace create GOV_STATUS
# 2. Create the D1 database for joke suggestions, paste its database_id into
# wrangler.jsonc, then apply the schema to the remote database
npx wrangler d1 create avemguvern-suggestions
npx wrangler d1 execute avemguvern-suggestions --remote --file=./schema.sql
# 3. Set the admin token (used to authorize POST /api/status and GET /api/jokes)
npx wrangler secret put ADMIN_TOKEN
# 4. Deploy
npx wrangler deploy
# 5. (optional) Seed the KV value — default already matches
npm run seed
For local development, apply the schema to the local D1 once:
npx wrangler d1 execute avemguvern-suggestions --local --file=./schema.sql
Custom domain
After the first deploy, attach avemguvern.ro in the Cloudflare dashboard
(Workers & Pages → avemguvern-ro → Settings → Domains & Routes), or add to
wrangler.jsonc:
"routes": [{ "pattern": "avemguvern.ro", "custom_domain": true }]
(DNS for the domain must be on Cloudflare.)
Updating the status
When the political situation changes, patch the status (only send fields that change):
curl -X POST https://avemguvern.ro/api/status \
-H "Authorization: Bearer $ADMIN_TOKEN" \
-H 'content-type: application/json' \
-d '{"hasGovernment": true, "answer": "Da!", "subtitle": "Avem guvern plin!", "interim": false, "primeMinister": "..."}'
updatedAt is stamped automatically.
Using the MCP server with Claude
Visit https://avemguvern.ro/mcp in a browser for setup instructions, or add it directly:
claude mcp add --transport http avemguvern https://avemguvern.ro/mcp
It exposes one read-only tool, get_government_status, that returns the current answer
plus the raw status JSON.
Abuse protection
Built into the Worker:
- Edge caching —
GET /api/statusis cached at the Cloudflare edge via the Cache API (caches.default) forCACHE_TTL_SECONDS(default 60), so a read flood hits KV at most ~once/minute per data center. A successfulPOSTpurges the cache so updates show immediately. The Cache API is per-colo, so a write purges only the data center that served it; other regions refresh when their TTL expires (hence the modest default — raise it inwrangler.jsoncvars for more offload if you can tolerate longer cross-region staleness after a change). - Per-IP rate limiting — native Workers rate-limit bindings: reads 120/min,
writes 10/min (the write limit also throttles token guessing). Over-limit →
429. - Hardened writes —
POSTonly accepts the known fields (hasGovernment,interim,answer,subtitle,primeMinister) with correct types, clamps strings to 200 chars, rejects bodies over 2 KB (413), and compares the admin token in constant time. A leaked token can't store arbitrary or huge data.
Recommended at the Cloudflare edge (dashboard):
- WAF Rate Limiting rule (e.g. per-IP threshold on
/api/*and/mcp) — this is the layer that drops abusive traffic before it counts against your quota. - Bot Fight Mode under Security → Bots.
Note: a zone-level Cache Rule cannot bypass the Worker here — for a Worker-owned route the Worker always executes before the cache is checked (docs). The in-Worker Cache API above is the effective edge-cache mechanism; it saves KV operations but each request still counts as one Worker invocation. To cut invocations themselves, use the WAF rate-limit rule.
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。