Skip to main content
Video tutorial · 1:46

Discovery inventory

The inventory is the union of every certificate discovered across your configured sources, de-duplicated by SHA-256 fingerprint. Filter, cross-reference with managed certs, and drill into the servers that serve them — the inventory is where you answer "what certs do we actually have?"

note

Scan results are persisted as they arrive, so during a long-running scan the inventory fills in progressively rather than in one burst when the scan finishes (refresh the page to see new rows). Anything that needs the whole scan still lands only when the run completes: run counters, disappearance detection, the findings refresh, new-certificate / certificate-changed notifications, and the cipher-suite enumeration pass — so a row discovered mid-scan shows an empty Accepted (weak) column until the run ends.

Default columns​

  • Certificate (the name it is known by) and Issuer.
  • Expires — relative time to not_after, colour-coded (red = expired, amber = within 30 days, green otherwise; the exact date shows on hover).
  • Key — algorithm and size.
  • PQC — Vulnerable, Transition, Ready, or Unknown. See PQC readiness.
  • Risk — the certificate's risk level, or OK. The list shows the level the findings refresh stored after the last scan (and every hour), so you can sort and filter by it; the expiry-based part can be up to about an hour behind. The certificate's detail page always computes it fresh. A certificate seen for the first time before its first refresh is scored on the spot.
  • Servers — how many servers (address and port) serve the certificate now. A server that answers for several names, or that several sources scan, counts once.
  • Tags · owner — what you set on the certificate's detail page, and in blue the tags its tag rules gave.
  • Status — Managed (links to the managed certificate), Older version (an earlier version of a managed certificate a server still serves — see Managed vs shadow) or Not managed.
  • Actions — Manage on a certificate that is not managed: opens the Manage wizard.
  • Under the name, No server has answered since <date> when every server that served the certificate failed its latest scans — see Servers that stop answering.

Click a column header (certificate, issuer, expiry, key size, risk) to sort.

Deprecated TLS versions​

Next to the negotiated TLS version, a server's page shows under Older versions whether the server also accepts the deprecated TLS 1.0 / 1.1 — the two are different questions, and a server that negotiates TLS 1.3 with you can still let a downgrade-capable client negotiate TLS 1.0.

  • TLS 1.0, TLS 1.1 — the server completed a handshake on that deprecated version. This is what raises the corresponding finding.
  • None accepted — no acceptance of either version was recorded. Normally that means both were refused, but the probe runs once per host and is not retried, so a host whose single probe could not complete its handshake also reads this way.

Answering this needs its own handshake, so it is probed once per host:port rather than once per name: TLS version support is a property of the listener, not of the server name you ask for. The measured answer is then shared with every endpoint row of that host, so all rows of one host agree. Rows discovered before this became the behaviour keep their old value until the next scan of that source.

Server history​

A certificate's Change history shows only the changes it took part in: the day it replaced another certificate on a server, the day another one replaced it, and TLS-settings changes while it was served. To see everything one address and port served over time, open the server's page: its History lists, newest first and across all its names,

  • every certificate replacement, with the certificate before and after (each links to its inventory page while it is still in the inventory; one discovery no longer sees is shown by fingerprint),
  • every TLS-settings change while the certificate stayed the same (TLS version, cipher, post-quantum key exchange, chain trust), with the values before and after.

Changes are kept for 365 days and the page shows the 100 newest. Changes recorded before 1.6.3 carry no endpoint, so they appear on the certificate's history but not on a server's.

Edge and origin​

A name behind a Cloudflare proxy (orange cloud) is discovered on both sides of the proxy by a Cloud DNS Provider source. In the server lists an edge tag marks the Cloudflare edge address the name resolves to on public resolvers — the certificate visitors see; the origin is the address in the DNS record, the certificate the origin server itself presents. The endpoint export and API carry the role (edge / origin) per row. Servers from records that are not proxied, and from every other source type, carry no tag.

A Name mismatch tag on a name — in the Names on this server table of a server's page — means the server answered a name that DNS points at that address with a certificate that does not cover it. Names the crawler only tried because they appeared in a certificate there, and that DNS does not point there, are never recorded. See Sources → Name mismatch.

Filters​

  • Search — matches subject, issuer and SANs.
  • Expiring within — 7, 30 or 90 days.
  • Issuer and Key type — the issuers and key types present, with how many certificates each has.
  • Managed — Managed (CertAutoPilot's current version), Older managed version or Not managed. The three never overlap.
  • Risk and PQC — one or more levels.
  • Source — certificates the chosen discovery sources currently see on a server.
  • Tag and Owner — see Tags and owner; they cover tags typed by hand and those given by tag rules.
  • Servers answering — A server answers or No server answers (every server that served the certificate failed its latest scans).
  • Trust — how the chain the servers send verifies on the servers that serve the certificate now: Public CA (the system's trusted roots — certificates from public CAs), Private CA (a CA added to the source's custom CA field — add the roots of your internal CAs there) or Not trusted. A certificate no server serves now (seen only in CT logs or on servers whose last scan failed) matches none of them.

The Overview page and Insights open the inventory with filters that have no control in the bar: an expiry range, a risk reason, a CAA verdict, a domain, no tag and no owner, self-signed and wildcard. They are shown under the bar as Also filtered by chips; close a chip to drop the filter. Picking Expiring within drops an expiry range and the other way round; picking a tag drops no tag, an owner drops no owner.

Filters, search, sort and the open tab (?tab=servers) live in the page address, so a filtered view can be bookmarked or shared; switching project clears them. The Managed / Not managed / Older managed version, Expiring within 30 days and PQC segments of the summary cards above the table show exactly that set when clicked (other filters and the search are cleared, so the list matches the count). Views saves the current filters under a name — see Saved views.

A user limited to granted discovery sources sees, filters and exports only the certificates its sources found.

Export​

Export → CSV or JSON downloads every certificate matching the current filters and search (not just the page): subject, SANs, issuer, validity, key, signature algorithm, PQC status, risk level, score and factors, the managed certificate's id, endpoint count (rows, not servers), the sources that see it, tags (typed and from rules), owner (typed, else a rule's), first / last seen, the SHA-256 fingerprint and, as the last column, Earlier Managed Version (whether it is an earlier version of a managed certificate). In the CSV a value a spreadsheet would treat as a formula starts with an apostrophe. An export that fails part-way is marked as incomplete (a last CSV line; JSON is left unterminated) so it is never taken for a complete one. Exports are recorded in the audit log and limited to ten per minute per project.

Endpoint export​

To see which certificate is served where, export the endpoints: one row per endpoint (address, port and name) with the certificate it serves. Export sits on every server list — the Inventory's Servers tab, the Servers page, a discovered certificate's Where it's served card (only that certificate's), a source's Servers tab (that source's) and the Quantum readiness list (the key-exchange view shown) — and takes the list's filters; failed rows are included when the list shows failed servers.

Columns: host, port, SNI, protocol (STARTTLS), source, status (managed / unmanaged / failed), certificate subject, issuer, expiry and SHA-256 fingerprint, TLS version, cipher suite, key-exchange group and class, chain status, edge / origin role, name mismatch, first and last seen, and the last error. A failed endpoint names no certificate: what discovery stored from its last successful scan is not what the address serves now, so the row carries the error instead. Failed endpoints are left out unless the screen shows them. The export is not capped, a user limited to some discovery sources exports only what those found, and it shares the ten-per-minute limit and the audit log with the other exports.

Tags and owner​

On a certificate's detail page, Tags and owner lets an operator add up to 20 tags (letters, digits, spaces and . _ : / -, stored in lower case) and a free-text owner — a team, a person or an email address. They are kept apart from the certificate itself: discovery deletes a certificate no endpoint serves any more, and when it is seen again its tags and owner are still there. The inventory and the findings list filter by them. Changes are audited (Discovered Certificate Tagged).

Tag rules can add tags and an owner on their own. What they give is shown under From tag rules on the same card and is kept apart: the rules never change the tags typed here, and an owner typed here wins over a rule's.

Cert detail​

Clicking a row opens the certificate's detail page:

  • Subject, SANs, issuer, validity, key and PQC details, plus the certificate checks: chain (trusted by the system or the source's CA, self-signed, untrusted root, intermediate missing, out of order), revocation (stapled or asked from the CA), the issuer URL (AIA) answer when the server sends an incomplete chain and the source fetches missing intermediates, extended key usage, CAA and lifetime.
  • Seen — N servers under M names: the servers (address and port) that serve it and the different names they answered for.
  • Where it's served — the server list filtered to this certificate, with the same filters, columns and export; click a server for its TLS settings, every name it answers for and its history. IPv6 addresses are shown bracketed ([fd00::1]:443). This replaces the earlier Endpoints and TLS grid.
  • Timeline — one lane per server that serves or served the certificate, a bar for each stretch it was served there and a mark for each TLS-settings change in between. A bar ends when the server switched to another certificate (named in the tooltip), when the name stopped answering, or is still open when the certificate is served now; amber means the server's last scan failed (the tooltip gives its last answer). A start older than the change history (365 days) is drawn faded and says start not recorded; when the axis starts at that point it is labelled first scan. While nothing has happened yet — every server still serves the certificate since it was first seen there, with no TLS change — the card says so in one sentence and lists the servers instead of drawing identical bars. Changes recorded before 1.6.3 cannot be put on a server and are only counted. At most 200 servers are drawn, the most recently seen first.
  • Change history — the changes this certificate took part in: the certificate a server served before it (or after it), and changes to the server's TLS settings while this certificate was served (TLS version, cipher, post-quantum key exchange, chain trust), each with the values before and after. For everything one server served, see Server history.
  • Open findings tied to this cert.
  • Tags and owner — see above.
  • Where it can go — the places the certificate was seen that can become distribution targets (its target candidates), each with the targets that already point there, switched between Candidates, Added and Removed (each with its count), with a search and 10 per page. It only shows them; to deploy, link the certificate to a module containing the target.
  • Manage with CertAutoPilot — promotes the cert to a managed certificate via the Manage wizard; on a cert that is already managed the button is Open managed certificate instead (Open the current version for an earlier version of a managed certificate).
  • Import with private key (project admins) — imports the certificate exactly as the server sent it, with its chain, as a managed imported certificate: you supply only the private key. See Import a discovered certificate.

Import a discovered certificate​

Discovery keeps the certificate and the chain each server presented (up to 12 certificates and 64 KB; a longer chain is not kept, the certificate alone still is). The import rebuilds the chain by following the certificate's issuers through what the server sent, leaving out expired certificates — such as an old cross-signed root kept for old clients — and anything that does not issue the certificate. Import with private key uses them with the key you paste or load from a file (and its passphrase if it is encrypted). The key must match the certificate and the certificate must not be expired; a certificate that is already managed (also one with the same fingerprint that a scan has not linked yet) is refused; the licence certificate limit applies, and the import is recorded in the audit log. The new certificate is linked to the discovered one and can be distributed like any imported certificate; it is not renewed automatically.

The button appears once a scan by this version has stored the certificate — a certificate only seen by earlier scans needs one more scan first. Certificates found only in CT logs cannot be imported this way (discovery never saw the server present them).

Managed vs shadow​

  • Managed: the cert is also in certificates — CertAutoPilot issued or explicitly imported it. You don't have to run the Manage wizard for a cert CertAutoPilot already issued: the scan that sees it automatically links a discovered cert to the managed certificate with the same fingerprint (also a partial scan, and also when another source of the project scanned a moment earlier), so a cert you issued here and then rediscovered flips from Not managed to Managed on its own. Deleting the managed certificate unlinks it again (the discovered record stays, as Not managed). The reverse view lives on the managed certificate itself: its detail page has a Discovery tab listing every live endpoint discovery has seen serving that certificate (its current leaf or a retained version), flagging any endpoint still on an older version after a renewal.
  • Older version (of a managed certificate): a server still serves an earlier version of a certificate CertAutoPilot manages — its current version was not deployed there. Discovery recognises it from the versions CertAutoPilot keeps of each certificate (the retention under Settings → General; a version pruned from that history reads as Not managed), whether it was seen before or after the renewal, and re-checks every hour and after every scan. It is marked only once a scan has seen it served after the current version was stored — a renewal whose distribution replaces it before the next scan never shows as an older version. It counts neither as managed nor as not managed, opens the managed certificate instead of the Manage wizard (managing it again would make a second certificate), raises an outdated_managed_version finding and gets the discovery expiry alerts. Once the servers serve the current version it is managed again — or leaves the inventory when nothing serves it.
  • Not managed (shadow): the cert exists in the wild but CertAutoPilot doesn't own its lifecycle. These certs are the top candidates for Manage import or for a conversation with the team that owns them. Once adopted through the Manage wizard the row shows Managed, and renewal and distribution flow through CertAutoPilot.

Servers that stop answering​

A server that fails a scan keeps the certificate it served when it last answered, so a brief outage does not empty the inventory. When every server that served a certificate fails its latest scan, the certificate shows No server has answered since <date> (the Servers answering filter finds them, and the Overview says how many expired certificates are only on such servers). A server that has failed every scan for 30 days counts as gone: at its source's next scan its rows disappear (with the usual endpoint gone notification) and a certificate nothing else serves leaves the inventory. The 30 days count from the first failed scan after its last answer, so a source paused for weeks does not lose a server to one failed scan after it is resumed; rows that failed before the upgrade start counting at their next failure. A server that answers again comes back on its own.

Stats widget​

At the top of the inventory, a summary card shows:

  • Certificates and Servers answering. Servers answering counts each address and port once, leaving out servers that failed at the last scan — the same number as the Servers tab and its list (hover it for the total with the failed ones; the API's total_servers includes them, total_endpoints counts the rows).
  • A Managed by CertAutoPilot bar — Managed, Not managed and, when there are any, Older managed version — and a Validity bar (More than 30 days, Expiring within 30 days, Expired). Clicking a managed part or Expiring within 30 days shows exactly that set.

Beside it, a Quantum readiness card has two bars:

  • Certificate keys — Vulnerable / Transition / Ready / Unknown.
  • Key exchange on servers — the post-quantum status of the negotiated key exchange across live servers (host:port, each counted once by its weakest endpoint row): Quantum-safe (ML-KEM) or Classical. Each part, and the card's Details link, opens Discovery → Quantum readiness.

The summary is cached for about 30 seconds, so a just-finished scan can take a moment to show in the tiles.

API​

GET /api/v1/projects/{projectId}/discovery/certificates
?expiring_within=30d
&self_signed=true
&wildcard=true
&search=example.com
&page=1
&page_size=100

expiring_within accepts 7d, 30d or 90d. managed accepts true (CertAutoPilot's current version), false (not managed) or outdated (an earlier version of a managed certificate); serving accepts answering or not_answering. expires_from / expires_to (RFC 3339 times, not_after in [from, to), at most 400 days apart, not together with expiring_within), risk_factor (stored reason codes), caa (ok, violation, unknown_issuer, error, unchecked), domain (a registrable domain such as example.com; matches certificates with a SAN or common name whose registrable domain it is — the Domains grouping), trust (public, private, untrusted), no_tag and no_owner are refused with 400 when invalid. Results are page-paginated (page, page_size up to 100) and return total. See API reference.

Retention​

Discovered certs are not kept forever: a certificate that no active endpoint serves and no server certificate scan location holds is deleted as an orphan — after a scan that fully covered its source, and in any case by an hourly cleanup. A certificate written in the last 5 minutes is never deleted (a scan of another source may be about to save the endpoint or location that holds it); deleting a source removes what only it held a few minutes later. So the inventory tracks what is currently out there, not a historical archive — record anything you need to keep for compliance before it stops being served.

See also​

Video tutorial

Short, hands-on videos showing this feature working for real. English audio, English and Turkish subtitles. All video tutorials →

Discovery Inventory — Filters, Saved Views, Export and ImportYouTube ↗