RSS Keyword Monitor

io.github.telemarkdigital-publisher由 GitHub 账号 telemarkdigital-publisher 发布v0.2.0更新于 Oct 27, 2026

Search bounded RSS, Atom, and RDF feed matches by keyword or regex.

已验证Streamable HTTP可网页运行效率与工作流网页搜索与抓取

概览

AI 生成的概览

远程 MCP 端点，运行 Apify RSS 监控任务，让助手在 RSS、Atom 和 RDF 订阅源中找出新的关键词、正则或排除规则匹配项。

可以做什么
  • 运行 Apify RSS 监控任务，按关键词、正则表达式和排除规则对 RSS、Atom、RDF 订阅源进行定时告警。
  • 默认只返回新匹配的订阅项，重复运行即可看到最新命中。
  • 可选将匹配结果通过 POST 发送到 webhook 目标。
适用场景
适合让助手持续盯住订阅源中的特定关键词、模式或排除条件，并只报告新增命中项，例如新闻、发布说明或安全公告的监控，而无需通读整个订阅源。
运行要求
需要一个远程 streamable HTTP 端点；未声明任何软件包、环境变量或请求头。README 说明底层工作流依赖 Apify 账号与 OAuth/MCP，导入工作流后需自行连接 Apify 和目标账号。
安装前请注意
该工具可将匹配结果 POST 到 webhook，数据会离开当前会话发送到你配置的目标。README 指出运行可能在写入数据集记录后以 FAILED、TIMED-OUT 或 ABORTED 结束，且不保证恰好一次投递；取回数量受已保存任务上限限制（maxItemsPerRun 最高 200）。示例中不含任何凭据。

安装

在 SourceWeft 中

  1. 打开 控制台中的 RSS Keyword Monitor，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.telemarkdigital-publisher%2Frss-keyword-monitor/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "rss-keyword-monitor": {
      "type": "http",
      "url": "https://telemark-apify-monitoring-mcp.katfu111111.workers.dev/rss"
    }
  }
}

README

Apify Monitoring Workflows

Public Apify Task configurations, validated n8n workflow examples, Make construction specifications, and live Make shared scenarios for monitoring Bluesky posts, RSS feeds, and TED procurement notices.

Created and maintained by Telemark Digital.

Start here

Products

  • Bluesky Keyword & Mention Alerts monitors public Bluesky posts for keywords, handles, mentions, and hashtags; its MCP route has public proof links for directories and AI agents: VerifyMCP trust and mcpbeat uptime.
  • RSS Keyword Monitor filters RSS, Atom, and RDF feeds by keyword, regular expression, and exclusion rules.
  • TED Tender Monitor is a TED tender monitor for new or changed EU procurement notices by CPV code, country, keyword, value, and notice type.

Each product directory contains public Apify Task inputs, a platform-neutral workflow contract, an n8n workflow, a Make implementation package, fixtures, and validation notes.

The TED package now includes thirteen live Apify Examples pages for high-intent procurement searches: AI procurement, cybersecurity, healthcare, SaaS/cloud software, construction, CPV 72 software, change notices and corrigenda, German managed detection, recent procurement JSON previews, EU software CPV previews, French medical equipment, Polish IT services, and renewable energy procurement.

The RSS package now includes eleven live Apify Examples pages: multi-feed keyword alerts, regex and exclusions, webhook-ready JSON, security advisories, vulnerability research, AI product news, research paper alerts, DevOps release notes, cloud platform updates, company news, and government cyber updates.

The Bluesky package now includes eleven live Apify Examples pages: keyword search, brand-handle mentions, hashtags, product launch keywords, developer tools launches, open-source maintainer discussion, AI policy posts, security keywords, data breach discussion, startup funding posts, and brand or competitor handle monitoring.

Live Make shared scenarios:

Live n8n workflow templates:

Live MCP and AI-agent directory surfaces:

AI client plugins

Foundational Codex, ChatGPT desktop, and Claude Code plugin packages are included under plugins. Each plugin exposes exactly one public Apify Actor through a product-specific Telemark MCP gateway route backed by Apify OAuth/MCP and includes a product-specific skill that keeps bounded one-off discovery separate from saved-Task monitoring.

The plugin packages passed repository-source installation plus live OAuth/MCP calls in Codex and Claude Code on 2026-07-23. These repository-source packages are separate from the official OpenAI and Anthropic review programs. See AI Client Plugin Marketplaces for the package layout, safe usage boundary, and current directory gates.

Remote MCP directory metadata for the same product routes is under mcp. Official MCP Registry, Smithery, and Glama listings are live for all three products. Smithery uses distinct free-capped product Worker URLs for cleaner tool discovery. Glama reports each connector as Healthy, VerifyMCP provides public trust pages, mcpbeat provides public uptime proof pages, and the public OAuth challenge, ownership, and server-card endpoints are live; see MCP Directory Listing Pack.

Public product icons are in assets/icons using clean external filenames. The AI-client plugin manifests omit optional icon fields for this release so the plugin packages remain text-only and easy to publish through GitHub's web editor.

Terminal-run reliability

Actor runs can commit dataset records before ending FAILED, TIMED-OUT, or ABORTED. For each terminal run they process, the public workflows fetch the run's defaultDatasetId when present, persist records under stable product-prefixed keys, and only then expose or report the terminal status. Failed empty datasets and runs without a dataset ID are reported safely. The terminal fixtures also prove replay idempotency, and the TED fixture confirms that title: null remains valid.

For Make, the construction specifications use Apify's maximum 1000-run task-runs page, one exact-key data-store cursor per Task, cursor-filtered reverse-page processing, first-run cursor priming, and an overflow stop whenever the fetched page does not contain the stored cursor run. Each publication template processes at most one selected Task run per scenario execution (maxRunsPerScenarioExecution = 1) so operation budgets stay bounded; if an account needs a longer outage window than 1000 * Apify Schedule interval, add paginated backfill before activation.

In each n8n export, dataset retrieval and Data Table upsert retry the same operation in place up to three times, waiting five seconds between attempts. These retries keep the original defaultDatasetId and stable delivery key and never rerun the Apify Task. If either operation still fails after its retries, perform manual recovery using the original run ID and its dataset; this package does not claim automatic exactly-once delivery.

Non-paginated delivery caps

The included n8n workflows and Make specifications intentionally make one non-paginated dataset request per Task run. Full retrieval therefore depends on the saved Task cap:

ProductSaved Task capFixed retrieval limitReserved control rows
BlueskymaxPostsPerRun <= 100100None
RSSmaxItemsPerRun <= 200200None
TEDmaxNewPerRun <= 9991000Exactly one appended summary row (999 + 1 = 1000)

Do not activate a workflow until the account owner has verified the matching saved Task setting. Raising a cap requires a separately designed, documented, and validated paginated workflow; it is outside this package.

Security

These examples contain no credentials. Connect your own Apify and destination accounts after importing a workflow. Read SECURITY.md before publishing or modifying an export.

Repository status

The examples are prepared and validated locally before publication. All three Make scenarios and all three n8n templates are publicly shared through the links above. The repository retains credential-free construction specifications rather than exported Make blueprint files.

See Account Gates for the GitHub publisher, n8n Creator Portal, optional n8n Cloud, and Make publication steps.

来源：README.md，提交 93066d4

工具

1

uplifted_novice_vbl--rss-keyword-monitor-only-new

RSS, Atom, and RDF keyword monitor

Run the Apify RSS monitor for scheduled keyword, regex, and exclusion alerts across RSS, Atom, and RDF feeds. Returns only new matching feed items by default and can POST to a webhook.

输入参数feedsonlyNewkeywordsresetStatewebhookUrldedupWindowmatchFieldsexcludeTermsregexPatternsincludeContentmaxItemsPerRun

版本历史

1
  1. v0.2.0最新Sep 16, 2026

评分

已满足适用检查项的 100%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分