international-jobs-mcp
Enables searching and retrieving UN, NGO, and international organization job postings via the ReliefWeb API.
README
international-jobs-mcp
A small MCP server that lets Claude search current UN, NGO, and international-organization job postings — UNDP, UNICEF, UNHCR, WFP, IOM, WHO, FAO, OCHA, and hundreds of NGOs/IGOs — via the official ReliefWeb API.
Why ReliefWeb, and not UN Careers / Impactpool / Devex?
Those sites don't publish a public API, and scraping them would be fragile and against most of their Terms of Service. ReliefWeb (run by UN OCHA) is the one major hub in this space that offers a free, official, read-only API — and it already aggregates postings from most UN agencies plus a huge number of NGOs, so it covers the same ground legitimately.
What it gives Claude
search_international_jobs— search/filter by keyword, country, organization, career category; open or expired postings; pagination.get_international_job_details— full description + how-to-apply for one posting.list_career_categories— current valid category filter values with live open-posting counts (e.g. "Programme/Project Management", "Information and Communications Technology", "Logistics/Procurement" — all relevant to your PM/ERP background).
⚠️ Before you do anything else: get an appname
Since Nov 2025, ReliefWeb requires a pre-approved appname for all API calls (it's free, just needs their sign-off):
- Go to https://reliefweb.int/contact (or find the "request an appname" form linked from https://apidoc.reliefweb.int/parameters).
- Describe it as a personal tool querying the Jobs API. Include an org/purpose string + random characters as your appname, per their instructions.
- Wait for their approval email — this is the one step only you can do, since it needs your contact details.
Step 1 — Smoke-test the API (do this first)
I wrote server.py against ReliefWeb's documented field-naming
conventions (country.name, source.name, etc.) but couldn't test it
against the live API from my sandbox — api.reliefweb.int isn't in my
reachable network. Once you have your appname, run this first:
pip install httpx
export RELIEFWEB_APPNAME="your-approved-appname"
python test_api.py
If any field name is wrong, ReliefWeb's error response names the bad
field directly — cross-check against
https://apidoc.reliefweb.int/fields-tables and fix it in server.py
(search for the fields.include lists — there are three of them).
Step 2 — Run it locally
pip install -r requirements.txt
export RELIEFWEB_APPNAME="your-approved-appname"
python server.py # stdio — for local testing with MCP Inspector
# or
python server.py --http # streamable HTTP on :8000 — for remote/Claude use
Test it with the official inspector before deploying anywhere:
npx @modelcontextprotocol/inspector
Step 3 — Deploy it somewhere with a public HTTPS URL
Claude's custom connectors need a reachable HTTPS URL, so pick one:
Option A — Render (fastest, free tier, good for a first pass)
- Push this folder to a GitHub repo.
- On render.com: New → Web Service → connect the repo.
- Environment: Docker (it'll pick up the included
Dockerfileautomatically). - Add environment variable
RELIEFWEB_APPNAME= your approved appname. - Deploy — Render gives you a URL like
https://your-app.onrender.com.
Option B — Railway (similarly quick)
railway initin this folder, thenrailway up.- Set
RELIEFWEB_APPNAMEin the Railway dashboard's Variables tab. - Railway gives you a public URL automatically.
Option C — Azure Container Apps (matches your RELAINE stack)
Since RELAINE already runs on Azure with GitHub Actions:
az group create --name international-jobs-mcp-rg --location eastus
az containerapp up \
--name international-jobs-mcp \
--resource-group international-jobs-mcp-rg \
--source . \
--target-port 8000 \
--ingress external \
--env-vars RELIEFWEB_APPNAME=your-approved-appname
You can wire this into the same GitHub Actions pattern you already use for RELAINE if you want CI/CD on every push.
Step 4 — Add it to Claude as a custom connector
In claude.ai: Settings → Connectors → Add custom connector, then paste
your deployed URL with the MCP path (check your server's startup log for
the exact mount path — FastMCP's streamable HTTP transport defaults to
/mcp, so it's typically https://your-app.onrender.com/mcp).
Once connected, you can ask me things like "search UN jobs in project
management, Tanzania or remote" and I'll call search_international_jobs
directly — same as the Jobs and Careers connector, but pointed at the
UN/NGO/international-org space instead of the US commercial market.
Notes / honesty check
- Rate limits: ReliefWeb caps this at 1000 calls/day and 1000 results per call — generous for personal use.
remote_onlyis a soft filter (adds remote/home-based keywords to the text search) — ReliefWeb has no hard structured "remote" flag like the US jobs tool does, so treat it as a nudge, not a guarantee.- If you later want UN Careers or Impactpool coverage specifically, that would require either an official partnership/API access from them, or scraping — which I'd steer you away from building, since it's fragile and against their terms.
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。