About ATX
ATX is an independent evidence and trust layer for AI-callable software. MCP servers are the first thing it measures, not the point of it.
What it does
It observes publicly reachable services without authenticating, records what it saw as immutable evidence, and derives an explainable score from that evidence using a versioned policy. Every published number links to the evidence behind it and to the query that produced it, because the alternative is asking you to take our word for it.
The scanner sends a handful of unauthenticated requests and, where a service speaks MCP, two read-only calls. It never invokes a tool. See the scanner page for exactly what it sends and how to make it stop.
Who runs it
ATX is built and operated by one person. There is no company behind it, no investor with an opinion about the scores, and no sales team. That is worth stating plainly because it cuts both ways: nobody can lean on the numbers, and there is no organisation to escalate to if you think one is wrong — just an email address and a commitment to correct the record in public.
How it is funded
It isn't, beyond hosting costs. No provider pays for a score, for placement, or for removal, and no score is influenced by payment — ever. If that changes, it will be disclosed on this page before it takes effect, not after.
What it will not do
- Invoke a tool on a service it is measuring.
- Present an inference as an observation. Findings derived rather than seen are labelled as inferences and capped in severity.
- Invent a fact to fill a gap. Missing evidence lowers confidence; it never becomes an assertion.
- Rewrite history. Evidence is append-only and enforced as such by the database — corrections are published as superseding records.
- Charge for a score, rank by payment, or sell placement.
If you think a finding is wrong
Say so. Prove control of the host and you can correct the record and trigger a rescan; the profile shows what we thought, when we changed it, and why. If a figure in the report is wrong, it is wrong in a way you can demonstrate from the published query — and we have already corrected ourselves in public once, which is documented in the report itself.
Opt-out requests do not require a justification and are not negotiated. Email admin@taladari.com or use robots.txt.