Zapier is great until the recipe that watches a public RSS feed and emails you also means Zapier stores the feed, your mailbox token, and a log of every event. IFTTT is the same deal with a friendlier coat of paint. If you want agents that scrape, filter, and notify, and you want the graph of those agents on hardware you control, Huginn is the project people have been running for that since 2013.
Huginn (huginn/huginn, ~50k GitHub stars, MIT, Ruby on Rails) is a self-hosted system of agents that create and consume events along a directed graph. Watch a site, an RSS feed, IMAP, a webhook, or MQTT; transform the event; then email, Slack, Telegram, SMS, or publish your own feed. The README calls it a hackable IFTTT or Zapier on your own server. That line still holds.
What agents actually do
You do not write a giant DAG YAML file. You add agents in the web UI, wire them together, and watch events flow. Typical jobs:
- Scrape a page and email you when the HTML changes (price, status, "applications closed").
- Follow RSS, IMAP, webhooks, and MQTT, then digest what matters once a day.
- Weather, flight, or deal watches with a human-readable alert.
- Peak/gap detection on an event stream so you notice silence, not only spikes.
- Post to Slack, Telegram, Pushover, Twilio, and similar from the last agent in the chain.
- Liquid, JavaScript, or jq for custom transforms; optional OpenAI-compatible APIs for summarize/classify if you want that layer.
- Write files to S3, FTP, or local disk; expose JSON or RSS of what your graph collected.
Complex or one-off agents can live as extra gems via ADDITIONAL_GEMS. Core stays general-purpose; weird company-specific scrapers do not have to land in the main repo.
Huginn vs n8n (honestly)
We already covered n8n. If you want a modern visual workflow builder, HTTP nodes, and a huge connector catalog, n8n is usually the easier hire. Huginn is older, more "small agents passing events," and still excellent at RSS, scraping, and "notify me when this public page changes" without paying per task. Plenty of people run both: Huginn for watchers, n8n for app-to-app glue. If you only have room for one and you live in OAuth SaaS APIs, start with n8n. If your problem is feeds and sites that will never ship a Zapier app, Huginn is the better default.
Why self-host?
The whole point of Huginn is that Zapier is not in the middle. Credentials, scraped HTML, email bodies, and location history (yes, there is a location agent) stay in your MySQL or PostgreSQL database. MIT license, no per-event bill. That matters for Canadian firms watching client sites or internal RSS without exporting the trail to a US automation vendor.
It is not a toy container, though. Production wants a web process plus a background worker, and a real database. The all-in-one image is fine to try; Compose with ghcr.io/huginn/huginn-single-process is what upstream recommends when you care about restarts. Default login on first Docker run is admin with a password printed once in the logs. Change that before the port faces the internet. Scraping other people's sites still has robots.txt and ToS constraints; Huginn does not make that go away.
Quick start
docker run -it -p 3000:3000 ghcr.io/huginn/huginnOpen port 3000, log in with the printed admin password, add a Website Agent or RSS Agent, and attach an Email or Slack Agent. For something you will keep:
# See github.com/huginn/huginn/blob/master/doc/docker/install.md
# Compose examples live under docker/single-process/ (MySQL or Postgres)Set APP_SECRET_TOKEN in .env, put HTTPS in front, and restrict outbound HTTP if the instance is shared (upstream documents request restrictions for multi-user boxes). Unraid and YunoHost packages exist; they can lag official images.
Who it is for
Good fit: journalists and researchers with a pile of RSS; homelabbers who outgrew a folder of curl cron jobs; agencies watching client status pages; anyone who liked IFTTT and hated the cloud copy of every trigger.
Skip it: if n8n already covers you; if you need a pretty new UI more than working agents; if you cannot run Rails + a worker + MySQL/Postgres without crying. Huginn's issue count is high because the project is old and ambitious, not because it is abandoned. Check recent commits (they are still landing) before you bet a company process on a dusty agent type.
Hosting in Canada
Huginn is happiest on a modest CloudBox VPS with persistent volumes for the database and enough RAM for MRI Ruby (do not starve it on a 512 MB toy). Keep the worker process alive; a web-only deploy will accept logins and never fire agents. Back up the DB. Put it in Canada if the events include client names, email, or anything you would not paste into Zapier.
Bottom line
Huginn is still the self-hosted answer when "little agents watching the web for me" is the spec. Fifty thousand stars is not hype from last week; it is a decade of people wiring RSS to email on their own metal. Start with Docker, replace one Zapier zap, and see if the graph clicks.
Talk with Swift Host if you want Huginn (or Huginn next to n8n) on Canadian infrastructure: TLS, database backups, and a worker that survives reboots. We are in St. Albert and actually answer the phone.
