Methodology

Two grades, never blended

Every reviewed service gets a Privacy grade and a Safety grade, each out of 10. They are never combined into one number on this site. Privacy answers "how little does this service learn about me?". Safety answers "how likely am I to get my money or my service?". A blended score lets a strong answer to the first question hide a weak answer to the second.

Each grade starts at 50 points. Points are added and subtracted by the fixed amounts below, the total is clamped to 0–100 and divided by ten. Every listing shows its own arithmetic line by line.

KYC levels

The KYC level is the largest single input to the Privacy grade. It is set from what a service's terms and documentation say, not from its marketing.

Level	Label	Meaning	Privacy points
0	ID never required	The terms rule out identity checks, or the design makes them impossible.	+20
1	No ID mentioned	No identity requirement appears anywhere in the terms or documentation.	+12
2	ID in rare cases	The terms allow identity checks only in exceptional cases, such as a legal order.	+4
3	ID on risk flag	The terms allow funds to be held and identity demanded after an automated risk flag.	-14
4	ID mandatory	Identity verification is required to use the service.	-32

Attributes

An attribute is a checkable fact about a service. Each one moves one or both grades by a fixed amount.

Attribute	What it means	Privacy	Safety
No registration	Usable without creating an account.	+6	0
Account-number login	Accounts are a random number or key. No email, no username.	+5	0
No email needed	An email address is not required.	+3	0
No phone needed	A phone number is never required.	+1	0
Tor onion service	Publishes an official .onion address.	+4	0
I2P address	Publishes an official I2P address.	+1	0
Works without JavaScript	Core functions work with scripts disabled.	+1	0
Accepts Monero	Monero is accepted or traded directly.	+4	0
Accepts Lightning	Bitcoin Lightning payments are accepted.	+1	0
Accepts cash	Cash by mail or in person is accepted.	+4	0
Open source	The code that matters is public and can be inspected.	+2	+8
Non-custodial	The operator never controls user funds.	+3	+12
Self-hostable	Can be run on your own server.	+3	+3
Independent audit	A third party has audited the code or the no-logs claim, and the report is public.	+2	+8
No-logs policy	States that connection or transaction logs are not kept.	+2	0
5+ years operating	Has operated continuously for five years or more.	0	+10
Legal entity disclosed	The operating company and its jurisdiction are published.	0	+5
Own liquidity	Trades from its own reserves, not an upstream exchange.	+1	+6
Multisig or hold-invoice escrow	Trades are protected by escrow the operator cannot spend alone.	0	+6
Refund without ID	Terms commit to refunding a stopped order without identity documents.	+2	+6
Runs an affiliate programme	Pays referrers. Shown for transparency; it does not change either grade.	0	0
Anonymous operator	Nobody is publicly accountable for the service.	0	-4
Excludes US users	Terms prohibit use from the United States.	0	0
Requires email	An email address is required to use the service.	-4	0
Requires phone number	A phone number is required to use the service.	-10	0
Custodial	The operator holds user funds, so it can freeze them.	-3	-5
Closed source	Code is not public.	-2	-3
Third-party liquidity	Orders are passed to upstream exchanges that apply their own screening.	-3	-6
AML screening	Terms mention blockchain analytics or risk scoring of deposits.	-7	-3
May request ID	Terms reserve the right to demand identity documents.	-7	-4

Test transactions

Editors run real transactions and publish each one on the listing: date, route, amount band, time to complete, whether ID was requested, and the result. Each successful test in the last 180 days adds +3 Safety points, up to +15. A test that is held, fails, or meets an ID request subtracts 10 Safety points, and an ID request during our own test also subtracts 10 Privacy points.

Small tests show that a service works. They do not show the size at which holds begin, which is why tests are paired with the incident registry.

Evidence tiers

Tier	Requirement
Unrated	Submitted by the community and not yet reviewed. No number is shown.
Docs reviewed	An editor has read the terms and documentation and recorded the facts. Grades are shown and labelled as documentation-based.
Tested	At least 3 successful published tests on 3 different days in the last 180 days.
Continuously tested	At least 6 successful published tests on 4 or more days in the last 180 days.

Tiers are calculated from the test ledger. A service drops a tier automatically when its tests age out.

Incidents

An incident is a documented event with a public source: held funds, a hack, a shutdown, added KYC, a data leak. Each published incident subtracts Safety points:

Severity	Base penalty
Low	−4
Medium	−10
High	−20
Critical	−30

The base penalty is multiplied by a factor for how the incident ended:

Outcome	Factor
Unresolved	×1
Funds lost	×1
Service closed	×1
Outcome unknown	×0.8
Resolved only after ID	×0.7
Resolved, user paid out	×0.35
Resolved	×0.35

Penalties fade in a straight line to a quarter of their size over 720 days, stay at that level for another year, then drop off.

Caps: rules that override the arithmetic

• An unresolved critical incident in the last 12 months caps Safety at 4.0.
• An unresolved high-severity incident in the last 12 months caps Safety at 6.0.
• A listing marked under dispute is capped at 4.0; one marked avoid at 2.0.
• A suspended service is capped at 3.0. A closed service has no grades.

Caps exist so that a service cannot keep a high Safety grade on the strength of old attributes while users are waiting for their money.

When a listing goes "under dispute"

A listing is marked under dispute when there is a documented, unresolved hold or loss of user funds above $5,000, supported by a transaction reference, and the service has not resolved it within 14 days of being asked. The service's public reply is shown alongside. The mark is removed when the case is resolved, and the history stays on the listing.

We describe what was reported and link to the source. We do not label a service a scam in our own voice.

User reviews

Reviews never change a grade. They are shown with the reviewer's account history, and reviews that include a transaction or order reference are marked. Reviews from new accounts are checked by a moderator before they appear.

The star average shown on a listing is trust-weighted. A review counts between one and six times depending on the reviewer's karma, and one and a half times more if it includes a transaction reference. That makes a wave of first-post five-star reviews worth very little.

Corrections

Anyone can send a correction from any listing. Operators can claim their listing by proving control of their domain and then reply publicly to incidents and reviews. Changes to a listing's KYC level, status or flags are recorded in its public history.

What is not listed

• Services under sanctions, and services seized by law enforcement, except as closed historical entries
• Mixing services, gambling and casinos
• Marketplaces for unlawful goods

Questions

Can a service pay for a better grade?

No. Grades are computed from the published formula on this page. Nobody, including the editors, can adjust a grade by hand, and payment of any kind does not enter the formula.

Why does a service with a high Privacy grade have a low Safety grade?

Because the two measure different things. A service can ask nothing about you and still hold funds, route through screened liquidity, or have an unresolved incident on record.

What does "Docs reviewed" mean?

An editor has read the terms, privacy policy and documentation and recorded what they say. No money has been sent through the service by us yet.