Compatibility

Indexer compatibility

indexarr does not contain a parser for every tracker: it speaks Torznab for torrents and Newznab for Usenet. Everything that exposes one of these protocols works without changes. For the rest, indexarr adopts the source with a native definition or with a family connector; if it has neither, the source stays queued and no proxy to Jackett or Prowlarr is created.

Matrix

Source	Status	Notes
Jackett / Prowlarr	not supported	indexarr replaces them: manager endpoints are rejected. Sources are added from the native catalog.
MirCrew (mircrew-indexer)	complete	Base or /api; magnet in link/enclosure; no seeders; /status for the forum login; slow responses.
Direct Torznab	complete (torrent)	Any tracker URL containing /api or torznab is used as is; otherwise /api is appended. Jackett/Prowlarr URLs are excluded.
Newznab / Usenet	complete (generic)	newznab connector and newznab-generic definition: caps, search/tv/movie/music/book search, .nzb payload stored or resolved at grab time, sending to SABnzbd/NZBGet clients. Native definitions for individual Usenet sites not yet in the catalog.
RSS / Atom	complete	Title, magnet, .torrent enclosure, size, numeric category, date; bulk import and OPML.
ExtTo	complete (HTML)	Dedicated parser: title and magnet from the same row; category from the c= filter.
TorrentGalaxy	complete (HTML)	Dedicated parser for post-detail pages with magnet cache (6 hours).
CorsaroBlu	listing only (HTML)	Dedicated parser for /torrent/, /torrents/ links and direct magnets, with detail cache. The public listing needs no credentials. Guest keyword search is disabled by the site; the CorsaroBlu (account) variant accepts a pasted session cookie and is still a draft. A FlareSolverr session is required because a plain client receives 429/503.
HTTP Basic	complete	Dedicated per-source fields, used for searches and downloads.
API key	complete	Sent as apikey and as the X-Api-Key header; it can be empty.
Cloudflare / challenge	with FlareSolverr	Automatic retry for RSS and direct Torznab; LAN addresses excluded.
Proprietary APIs	not handled automatically	The source stays in the "to be written" queue until a native definition exists; it does not appear in the user list. No automatic proxy to Jackett/Prowlarr.

MirCrew

mircrew-indexer is the reference case of a "hand-rolled" Torznab service: it deserves an explicit list of what indexarr does for it.

MirCrew behavior	indexarr handling
Endpoint <base>/api, port chosen by the user	Accepts both the base and the URL with /api
Magnet in <link> and <enclosure>	Recognized, no attempt to download it over HTTP
No seeders attribute	Seeders unknown (-1), attribute omitted in the output
<response offset total> without namespace	Ignored, pagination uses the request parameters
Date = forum topic date	Read from <pubDate>
/status with logged_in, login_error, last_run	The Test button queries it and distinguishes failed, in-progress and lost login
Slow forum search	Default timeout 45 s, run with margin
XML with raw & or HTML entities	Sanitization before parsing

MirCrew configuration

Field	Value
Name	MirCrew
Type	Torznab / indexer
URL	http://<host>:<port> or http://<host>:<port>/api
Manager	MirCrew (or Direct Torznab, Auto is fine)
API key	the service's own, if present
Categories	5000 for TV, 2000 for movies
HTTP username/password	only if the service requires them

The path has been verified against a real mircrew-indexer (version 0.1): caps, /status with an active forum login, ingestion of 100 releases from the "latest" list and Torznab search with magnet and infohash.

Diagnostics

Symptom	Meaning
Test: "forum login failed: …"	MirCrew forum credentials must be redone in the service; the local index keeps responding
Test: "forum login in progress"	Service just started, not an error
Test: "forum login lost"	The service had completed a run and lost the login
FlareSolverr test failed	Service unreachable, wrong URL or no Internet access from the solver
HTTP 401	Wrong API key or Basic credentials
HTTP 404 on download	Release stored only as a magnet: use the magnet link
HTML instead of XML	The indexer returned a login page or a block: check the URL and credentials
No results with an empty query	Direct Torznab may not support the "latest" list: set a search term on the source

HTML listings (ExtTo, TorrentGalaxy, CorsaroBlu)

extto.org/ext.to, torrentgalaxy.* and CorsaroBlu are not feeds: they are HTML listings. indexarr recognizes them by host and uses the same approach as Gextto:

Site	How
ExtTo	Pairs the post-detail link and the magnet present in the same row; no detail fetch.
TorrentGalaxy	Follows the post-detail links of the rows, extracts the magnet from the page and caches it for 6 hours.
CorsaroBlu	Follows /torrent/ or /torrents/ links and direct magnets, extracts magnets in parallel and caches them. It is a latest-list source; guest search is disabled and all requests use FlareSolverr.

Limits: first page, at most 50 releases per source; if the layout changes or the site requires login, the test reports the error. Quick add recognizes the host; Default category assigns a Torznab category when the page does not provide one. The other sites without RSS remain in the native queue: the index classifies them as "to be written" (maintenance view /indexers/new?queue=1) and does not create a Torznab endpoint towards the manager.

Cloudflare

With an endpoint configured in Settings → Cloudflare / FlareSolverr, a response with a Cloudflare challenge is retried through the service:

• cf-mitigated: challenge header, or HTML page with Just a moment..., cf-chl, __cf_chl, Attention Required, on 403/503 (and on 200);
• RSS, direct Torznab and connection tests;
• no retry for LAN addresses: a local 403 is an authentication/network error, not a challenge;
• an XML/JSON error from the indexer is not treated as a challenge.

Limits: .torrent files are not downloaded through FlareSolverr (binary response) and a resolution can take up to ~60 seconds. The indexer URL, including the API key, is sent to the FlareSolverr service.

Adding an indexer that is not listed

1. If it exposes Torznab, create a source with the full URL and the API key (if present). The Direct Torznab type avoids any heuristic about the manager.
2. If it requires HTTP Basic, fill in the dedicated fields.
3. If it only speaks its own proprietary API, the source stays queued: write a native definition by observing the site. indexarr does not create a proxy to Jackett/Prowlarr, does not contact them and does not implement ad-hoc protocols.
4. If the service exposes a /status next to /api, the test reads it automatically when the JSON contains ok or logged_in.

To verify without waiting for the scheduler, use Test or Run on the Sources page and check Runs for the exact error returned by the indexer. Quick add recognizes the type from the URL; Discover feeds finds the feeds declared by a page; presets set type and manager without constraining the URL.

Newznab / Usenet

The Usenet path uses the generic newznab connector and the newznab-generic definition. It is not a native definition for a single site: the URL and API key are provided by the user, so the catalog does not yet contain Usenet sources verified site by site.

Step	Behavior
Adding	API URL (with or without /api) and API key; the protocol is usenet and is derived from the definition
Caps	t=caps read and cached for an hour; the declared modes are search, tv-search, movie-search, music-search, book-search
Search	Newznab parameters (q, season, ep, imdbid, tvdbid, tmdbid, rid, artist, album, author, title, cat); entries without a valid link are discarded
Payload	.nzb downloaded with t=get or from the application/x-nzb enclosure; if the download budget is exhausted, a sealed reference remains, opened at grab time
Output	Torznab/Newznab with protocol=usenet, application/x-nzb enclosure and .nzb file name; no magnet and no infohash
Apps	Sonarr, Radarr, Lidarr, Readarr and Whisparr receive Newznab/NewznabSettings with protocol usenet; eD2k remains excluded
Download	the release is sent to the Usenet client (SABnzbd or NZBGet) configured for the usenet protocol

Measured limits: live search uses the same fan-out as the other protocols and does not download the NZB in advance; the payload arrives at the first grab. Native definitions for individual Usenet sites are not part of the current catalog, so a source that does not expose Newznab stays queued.

Download clients

Sending to clients is not part of the Torznab protocol: it is an optional action of the release detail, configured in App → Download clients.

Client	Integration
qBittorrent	Web API v2: login, torrents/add with magnet or .torrent file, category, path, pause.
Transmission	RPC /transmission/rpc: torrent-add, base64 metainfo for .torrent files, filename for magnets, 409 session handled.
Deluge	JSON-RPC: login, core.add_torrent_magnet/core.add_torrent_file, path and pause.
rTorrent	XML-RPC: load.raw_start/load.start with magnet or .torrent.
SABnzbd / NZBGet	Usenet API: add a .nzb or a URL, category, pause.
aMule	amuleapi Web API (/api/v1): password login, POST /downloads with the ed2k:// link. Protocol ed2k.
Aria2	JSON-RPC /jsonrpc: aria2.addUri and aria2.addTorrent, token, dir, pause.
uTorrent	Web API: token from token.html, add-url and add-file.
Flood	REST /api: authentication, adding magnet/URL/file, tags, destination.
Download Station	Synology auth.cgi + task.cgi: magnet/URL and file upload, destination.
NZBVortex	/api API: nonce login, add .nzb, group.
Hadouken	JSON-RPC /api: webui.addTorrent with file or URL, label.
Blackhole	Local folder: writes .torrent, .magnet or .nzb; no URL required.

Sonarr/Radarr/Lidarr continue to manage the download on their own; this feature is for sending a release chosen manually from the UI.