Kagi is a paid, ad-free search engine operated by Kagi Inc., a Delaware Public Benefit Corporation founded in 2018 by Vladimir Prelovac in Palo Alto, CA. Its business model is subscriptions only: no ads, no analytics or telemetry, and no tracking of result clicks. An account (email) and payment method are required; Kagi states search queries are logged only temporarily for debugging, purged quickly, and not linked to accounts, and payment data is not used to link users to activity. Each query triggers anonymized API calls to roughly a dozen upstream sources, including major search providers (the founder has confirmed Yandex among them) and, for AI features, third-party LLM providers with retention of up to 30 days. A 2025 Privacy Pass option provides cryptographically unlinkable searches. Kagi publishes a warrant canary and an independent security audit by Illumant.
TIMELINE
NO ENTRIES ON FILE YET.
TOOLS WE RECOMMEND
PLUS
RECEIPTS
18 ON FILE
[2025-01-11] · COLLECTION
As assessed by PrivacySpy on 2025-01-11 (see PrivacySpy's rubric entry for this product for the per-claim citation): Is it clear why the service collects the personal data that it does? — Yes.
Kagi's Privacy Pass announcement (2025-02-13) acknowledges that outside Privacy Pass mode, searching Kagi requires being logged into a paid account, making searches theoretically linkable to that account (while stating Kagi does not link searches to accounts or permanently record them).
Kagi's privacy policy (changelog entry 2026-02-17) states that while a record of usage volume is kept for billing, search queries are only logged temporarily for debugging and automatically purged, that server request logs are retained about 7 days and not linked to accounts, and that users seeking provable anonymity may sign up with a pseudonymous email, pay with cryptocurrency, use Tor, or authenticate via Privacy Pass.
Kagi's company page states Kagi was founded in 2018 by Vladimir Prelovac in Palo Alto, CA, was bootstrapped from 2018 to 2023 with about $3M from the founder, raised about $2.5M in 2023–2024 from roughly 93 angel investors drawn from its user community, became a Public Benefit Corporation in early 2024, and publishes real-time growth and usage statistics at kagi.com/stats.
Kagi's blog post of 2023-06-28 discloses its first external fundraise: a $670K SAFE note round from 42 accredited investors, most of them Kagi users, framed explicitly as an alternative to ad-tech-aligned funding.
As assessed by PrivacySpy on 2025-01-11 (see PrivacySpy's rubric entry for this product for the per-claim citation): Is the policy's history made available? — Yes, with revisions or a changelog.
Kagi's LLM privacy documentation lists the third-party providers behind its AI features — including Anthropic via Amazon Bedrock, OpenAI via Microsoft Azure, Mistral AI, Google Vertex/Gemini, Cerebras, Groq and others — stating none train on user data sent via API, with retention ranging from no storage to 30 days (Anthropic flagged prompts: 2 years).
Kagi's search-sources documentation states that, in addition to its own Teclis and TinyGem indexes, results include anonymized API calls to all major search result providers worldwide plus specialized sources (Marginalia, Wolfram Alpha, Apple, Wikipedia, Open Meteo, Yelp, TripAdvisor), with a typical query calling about a dozen sources simultaneously; the page does not name the major providers.
As assessed by PrivacySpy on 2025-01-11 (see PrivacySpy's rubric entry for this product for the per-claim citation): Does the policy allow personally-targeted or behavioral marketing? — No.
In a July 2025 Hacker News comment, Kagi founder Vladimir Prelovac (as 'freediver') confirmed Kagi has used Yandex's search API since 2019 and that Yandex represents about 2% of Kagi's costs, defending the relationship on technical merit after recurring community criticism.
Kagi's security documentation states its security was independently audited by Illumant, with the audit finding Kagi 'Highly Secure' with no findings of material significance, and that Kagi operates a bug bounty program with a safe-harbor policy for researchers.
Kagi's bug bounty documentation publishes program scope, a severity-to-reward matrix (high severity USD 500-1,000, moderate USD 100-500, low USD 50-100), an intake and adjudication process, and a running table of rewarded reports naming each finding, its severity, the amount paid and the date — for example USD 300 for arbitrary JavaScript execution in Orion Browser (2025-04-03) and USD 150 for improper access control on shared Kagi Assistant threads (2025-06-04). The page also states no legal action will be initiated against researchers who comply with the policy.
Kagi also publishes the earlier, less favourable stage of the same engagement: Illumant's summary report dated 2022-05-23 rated Kagi's application 'Reasonably Secure' rather than 'Highly Secure', noting security issues that deserved consideration, and the rating moved to 'Highly Secure' in the 2022-08-26 report after remediation. Publishing both the pre- and post-remediation reports is documented evidence of a fixed finding rather than a single favourable summary.
Kagi publishes the full Illumant audit report rather than only describing it: the Web Application Security Assessment Executive Summary Report dated 2022-08-26, signed by Illumant Director of Advisory Services Matija Siljak, records credentialed and uncredentialed penetration testing of Kagi's web application, assigns the rating 'Highly Secure' (defined by Illumant as no findings of material significance), and states the findings identified were of very low severity.
Kagi's privacy policy contains a warrant canary stating that customer information has never been disclosed, seized, or breached, and that Kagi has received 0 National Security Letters, 0 gag orders, and 0 warrants from any government organization; this is a self-attested statement, not third-party verified.
In February 2025 Kagi deployed Privacy Pass authentication, a cryptographic protocol under which Kagi cannot link redeemed tokens to a token-generation event or to each other, so it cannot tell which subscriber performed a given search; Kagi's own post notes IP addresses and browser fingerprints remain as side channels and recommends Tor for stronger guarantees.
Kagi's privacy policy states the service loads no analytics or telemetry, does not track which search results users click, and proxies media results and Kagi Assistant connections through its own servers.