SearXNG is a free, open-source, self-hostable metasearch engine — a 2021 fork of searx — that forwards a user's query to up to 270 upstream search services (Google, Bing, DuckDuckGo, Startpage, Qwant and others) and aggregates the results. Its own documentation states that users are neither tracked nor profiled, that no cookies are sent to the upstream engines and a random browser profile is generated per request, and that the software serves no ads or tracking content; no account exists and JavaScript and cookies are optional. The project is equally candid that a public instance's administrator sees the user's IP and queries and could log, aggregate or sell them, that queries reach the upstream engines from the instance's IP, and that users should therefore self-host or use an operator they trust; the project-run list searx.space grades public instances and flags those that add analytics or third-party files. The code is published under AGPLv3 on GitHub, maintained by a named lead (Markus Heiser) and an open volunteer community with no legal entity or disclosed funding beyond in-kind service plans; no security incident, breach or audit involving the project was found in this research.
TIMELINE
NO ENTRIES ON FILE YET.
TOOLS WE RECOMMEND
PLUS
RECEIPTS
16 ON FILE
COLLECTION
The Limiter documentation describes optional IP-based bot protection: to keep upstream engines from CAPTCHA-blocking the instance, an operator can enable rate limiting that determines each client's IP (via X-Forwarded-For / X-Real-IP), keeps IP block/pass lists and dynamic rate-limit state in a Valkey database — meaning an instance that enables it processes and temporarily stores users' IP addresses.
The 'general:' settings reference shows the software's defaults: enable_metrics is on by default and records anonymous engine metrics exposed at /stats, an operator may add a privacy-policy link (privacypolicy_url, off by default), and a debug switch activates verbose logging that the docs warn is intended only for local development — i.e. what an instance records is an operator-controlled setting, not fixed by the software.
The project's 'Why use a private instance?' page explains the anonymisation mechanism: SearXNG sends no cookies to the external search engines and generates a random browser profile for every request, so the request is anonymised whether the instance is public or private; the IP address the upstream engines see is the instance's own, and the instance can be configured to route through a proxy or Tor.
The SearXNG documentation index describes the software as a free internet metasearch engine aggregating results from up to 270 search services and states that users are neither tracked nor profiled; its feature list adds that JavaScript and cookies are optional and that it can be used over Tor. There is no account or sign-up in the software.
The searxng/searxng repository on GitHub — the project's public source and issue tracker — shows about 35.6k stars, 170 watchers and 3.2k forks, an AGPL-3.0 licence badge, a contributing guide and a security policy, evidencing a large, active and inspectable community project rather than a closed vendor.
The documentation index states that SearXNG development began in mid-2021 as a fork of searx, that the project is 'driven by an open community', and — under Acknowledgements — that the only disclosed material support is organisations giving the project free access to their paid plans; the site is copyrighted to 'SearXNG team' with no legal entity, governance body or funding figures published.
SearXNG's complete source code is published under the GNU Affero General Public License version 3, so anyone can inspect, audit, fork or self-host the exact code that runs an instance — and network operators who modify it are obliged to publish their changes.
The repository's AUTHORS file states that SearXNG is a fork of searx maintained by Markus Heiser (@return42) and lists dozens of named and pseudonymous contributors, including original searx author Adam Tauber; no company, foundation or registered association is named.
Privacy Guides lists SearXNG on its Search Engines page under 'Metasearch Engines' (as an actively maintained fork of searx), while cautioning that search queries are still sent to the upstream engines SearXNG aggregates from, that users should read the privacy policy of whichever instance they use, and that instances may be modified by their owners so their behaviour need not reflect that policy.
The project's own documentation states that a user of a public instance must trust its administrator, because they cannot know whether their requests are logged, aggregated, and sent or sold to a third party — the operator of whichever instance a user picks sees their IP and queries — and recommends running a private instance or using one operated by people the user trusts.
The project's About page states that SearXNG does not build a profile of the user, 'never shares anything with a third-party', and works by mixing a user's queries with other users' searches on the upstream platforms without storing search data — a description of the software's own storage that sits alongside the docs' separate acknowledgement that queries do reach third-party engines and that instance operators may log.
The repository's SECURITY.md publishes a responsible-disclosure contact (security@searxng.org) and asks reporters for reproduction steps; it describes no bug bounty, no third-party audit and no transparency report, and no breach or regulatory action involving the SearXNG project was found in this research — a clean record without affirmative artifacts.
The project's own instance list warns that public instances, because they relay high traffic from a single IP, have a higher chance of being blocked by upstream providers such as Google, Qwant, Bing and Startpage and may therefore yield less accurate results — a documented reliability limitation of the public-instance model rather than a privacy incident.
The project-run public instance list searx.space grades each listed instance on TLS, CSP and an 'HTML grade' that checks whether the served scripts match the SearXNG repository; grades range from V (vanilla) through C/Cjs (modified files) to E (includes files from another domain), and an eye symbol marks instances that run analytics software — the page itself cautions that only inline and external scripts are checked and 'there are many other ways to leak information'. Some public instances therefore add tracking the software does not ship.
The 'server:' settings reference shows the shipped defaults include an image_proxy option (so image results can be fetched by the instance rather than the user's browser), a public_instance flag, a Limiter switch, and default HTTP response headers of X-Robots-Tag noindex,nofollow and Referrer-Policy no-referrer; a POST search method is offered so queries are not saved in browser history.
The project's privacy page states that SearXNG does not serve ads or tracking content, forwards nothing from third-party services (such as advertisement) through the search engines, and hides both the referring page and the search query from the result pages a user visits.