Jackett

Last reviewed Sep 13, 2026 · 779 words

Jackett's job is translation. Sonarr asks for episode 4 of a show in Torznab, Jackett turns that into a search on any of several hundred torrent sites, and hands the results back as an RSS-shaped feed the arr apps understand. It did that job alone for years. Now that Prowlarr has matured, the default answer for a new stack is Prowlarr, and Jackett's remaining case is narrow but real: a tracker with a Jackett definition that Prowlarr lacks, or a piece of software that wants one aggregate Torznab URL and nothing else.

Why Prowlarr won

Prowlarr does what Jackett does and then syncs each indexer to every arr app for you. With Jackett you configure a tracker once in Jackett, then paste its Torznab URL and API key into Sonarr, again into Radarr, again into Lidarr, and repeat for every tracker you add. Ten trackers and three apps is 30 manual entries, and when a tracker changes its URL you edit it in all three places. Prowlarr turns that into one entry per tracker with a sync to every app, keeps statistics on which indexers actually return grabs, and handles the same Cloudflare workarounds. The definitions overlap heavily because Prowlarr's team came out of the same tradition. That is the whole case, and it is decisive for anyone building the arr stack from scratch.

The aggregate endpoint is Jackett's real trick

Jackett exposes a combined indexer at /api/v2.0/indexers/all/results/torznab/ that searches every configured tracker at once and returns one merged result set. Prowlarr deliberately does not offer this; its authors consider it bad practice because one slow tracker holds up the whole search, and the arr apps handle per-indexer settings better anyway. But for software outside the arr family, a media manager that accepts exactly one Torznab URL, or a script that wants search-everything in one request, the aggregate is convenient and nothing else provides it. Jackett also carries a handful of tracker definitions that have never been ported, usually small private or regional sites. If your must-have tracker is on that list, you run Jackett for it, and you can add Jackett's feed to Prowlarr as a generic Torznab indexer so the rest of the stack stays tidy.

Running it

It is a .NET application, about 200 MB of RAM at rest, with a web UI on port 9117. Native packages exist for Windows, macOS and Linux, and the linuxserver container is the usual route:

services:
  jackett:
    image: lscr.io/linuxserver/jackett:latest
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Etc/UTC
    volumes:
      - ./jackett-config:/config
    ports:
      - "9117:9117"
    restart: unless-stopped

Set an admin password in the UI before anything else, because the default is none and the config page shows every tracker credential you enter. Keep it off the public internet entirely; nothing but Sonarr and Radarr on the same network has any reason to reach it.

Trackers break, and the YAML definitions are why it recovers

A torrent site changes its HTML and every scraper that parsed it fails the same afternoon. Jackett's answer is Cardigann-style YAML: most of its several hundred trackers are described in a text file that says where the search form is and how to read the results table, and the project ships updated definitions constantly, so an update usually fixes a broken tracker within days. The cost is that updating Jackett becomes a weekly ritual, and a tracker can be dead for a week with nothing to do but wait or edit the YAML yourself. That churn is identical in Prowlarr, and it is the honest price of using public trackers at all.

Cloudflare walls need a helper

Many trackers sit behind Cloudflare's browser check, which a .NET HTTP client cannot pass. Jackett handles this by handing the request to FlareSolverr, a headless Chrome service, configured once under Jackett's settings as the FlareSolverr API URL. Budget another 512 MB or more of RAM for it and expect it to be the flakiest piece in the chain.

What I'd do

For a new media stack: Prowlarr, not Jackett. For an existing Jackett install that works: leave it until a tracker you care about breaks, then migrate, because there is no export path and re-adding indexers is a manual evening either way. I would only run both when a single tracker or the aggregate endpoint forces it, and then I would feed Jackett into Prowlarr rather than into each arr app directly, so there is still exactly one place where indexers live.