The observation layer
for AI agents.
Point it at any website and it returns what is actually there — certificates, DNS records, security headers, structured data, accessibility — as 200 deterministic checks, each carrying the evidence behind it.
POST /mcp
GET /api/v1/summary
Evidence on every finding
Models are good at reasoning.
They are bad at knowing what is true.
You described a site and a model built it. It renders, so it looks
finished — but almost nothing that decides whether it actually works is visible in a
browser preview. Whether the canonical points at the right URL. Whether a stray
robots.txt rule is keeping the page out of search entirely. Whether the
certificate chain is complete for visitors who are not you. Ask the model and it will
answer confidently from the code it wrote, which is not the same thing as what is live.
Outrings goes and looks. Observation and interpretation are not the
same thing, so they are not in the same fields: what is there — tested,
found, evidence — is measured, and what to make of it —
why, fix, severity — is our reading, marked as
such. An agent that would rather draw its own conclusions can ignore the second set
entirely, which is the whole reason the two are kept apart.
- You prompt “Build me a landing page for my SaaS.” · “Check my site before I launch it.” · “My site isn’t getting any traffic — what’s wrong with it?”
- The agent needs facts Often it is being asked to grade its own homework: the model that generated the site is now judging it. It can reason perfectly well, but it cannot look — so it answers from the code in the conversation rather than from what is deployed, and a confident guess reads exactly like a measurement.
-
Outrings goes and looks
One call,
RESTorMCP— no key, no account. The deployed site is fetched live and put through 200 checks across 18 categories, including the ones nobody thinks to look at until launch day.- SEO & structured data
- Security headers
- TLS & certificates
- DNS & email auth
- AI crawler access
- Infrastructure
- Accessibility
- Privacy & policies
- Content & credibility
- and nine more →
- It fixes what is actually broken Each finding comes back with the header, field or directive it was read from, so the agent repairs the real problem instead of rewriting code that was already fine — and anything that could not be determined comes back marked undetermined rather than quietly guessed.
What makes this different
Four things: every finding arrives with the observation it came from, anything undetermined is reported as such rather than counted as a pass, the same inputs always produce the same result, and the whole thing is built to be called by a machine rather than read off a dashboard.
Evidence, not assertion
Every finding ships with the raw observation behind it — the header, the certificate field, the directive that decided the verdict. A model quoting Outrings can show its working instead of asking to be believed.
Uncertainty is reported
What cannot be determined is returned as unknown, blocked,
timeout or requires_browser — never quietly scored as a pass.
Each response lists what was not checked, so an agent cannot over-claim from it.
Reproducible by construction
Each check is a direct observation plus a rule written by hand. No third-party scanners, no external APIs, no language model anywhere in the scoring path — the same observations under the same ruleset produce the same result. See all 200 checks.
Built for agents, not dashboards
A native MCP server and a keyless REST API return structured findings an agent can act on. The human interface is a view onto the same data, not the other way round.
Prove it to your visitors
Put a verified badge in your footer showing your security or overall score. It links through to a live audit, so the claim is checkable rather than decorative — and it updates as your score does.
Build a badgeGive your agent a tool
Point any MCP client at
/mcp and your assistant stops guessing about websites. One call returns a
measured verdict with evidence, confidence and an explicit list of what was not checked.
GET /api/v1/summary?url=example.comOutrings for AI agents
Questions
Is this generated by AI?
No. Every check is server-side code — a direct observation plus a rule written by hand — and no language model is involved in producing any score. The guarantee is narrow and precise: the same observations under the same ruleset produce the same result. If your certificate, DNS or headers change, the observation changes and so does the score. Every check is documented.
What can an AI agent do with it?
Connect a Model Context Protocol client to /mcp and eight read-only tools arrive already described, so the model decides when measuring beats guessing. Or call GET /api/v1/summary?url=… directly — no key, no account. See what the tools return.
What does it measure?
200 checks across 18 categories: SEO, AI crawler access, security headers, TLS, DNS and email authentication, accessibility, privacy, content quality, infrastructure and more. Each finding carries the observation it came from — the header value, the certificate field, the robots.txt line that decided it.
What does it not measure?
Stated on every response rather than buried: no WHOIS or domain age, no CVE matching, no PageSpeed or field Core Web Vitals, no backlinks or rankings, and nothing that needs JavaScript to run. DNSSEC and DKIM are reported as untested with the reason. Anything undetermined is returned as such and never scored as a pass.
Is it free?
Yes. No account, no cookie, no API key and no paid tier. Requests are rate-limited to 40 audits an hour so the sites being measured are not burdened. Terms.
Why Outrings
An inspection starts at one address and works outward. The page itself, then the headers around it, then the certificate, the DNS records, the mail authentication, the crawler directives, the third parties it pulls in — each ring further from the origin, each one measured rather than assumed.
What comes back is not a grade. It is a set of observations about a domain and everything it touches, structured so a machine can reason from it. That is the layer we are building: the outer rings of what is knowable about a web property, made queryable.