Skip to content

API Reference

This page is the technical reference for the IP Intelligence API: base URL, authentication, request and response schemas, every stable error code, and pagination and rate-limit mechanics.

It assumes you have already read Using the API and have a key. The query syntax — supported search fields, combination rules, and what each one searches — is defined once, in Supported searches; this page does not repeat it, only how to send and read it over HTTP.

Base URL

https://api.ip.alphastrike.io

Every route below is under /v1. There is currently one version; a route or field is not removed or repurposed within it.

Authentication

Send your key as a bearer credential on every request:

Authorization: Bearer ipk_live_7pQ2m9XvB4hK1sR6tY0wZ3nC8dF5gJaL

A request with no Authorization header, a malformed one, or a key that is unknown or has been revoked all answer the same 401 unauthorized — the response does not distinguish which, so it cannot be used to test whether a particular key string is valid. See Creating a key and Allowance, revocation, and security for how keys are created, revoked, and should be handled.

Using the API is permitted under § 5 (4) of the terms of service, with a credential issued for your own account, inside the same daily allowance the browser draws on, and within the rate limits Alpha Strike sets. Bulk retrieval and systematic extraction beyond that remain prohibited, and § 10 (2) still governs what you may do with results once you have them.

Request conventions

  • Send Content-Type: application/json with every request body. GET routes take no body.
  • A body must be valid JSON, at most 512 KB, and contain only the fields documented for that request. An unrecognized field is rejected rather than silently ignored — build your client to send only the fields shown below, so a future optional response field does not require a client change.
  • Every response — success or error — carries a request_id string in the body and the same value in the X-Request-ID response header. Include it when reporting a problem.
  • Timestamps are RFC 3339, always UTC (a trailing Z, never a UTC offset).

Rate limits and the daily allowance

Only POST /v1/search draws on the daily allowance described in How much you can search. GET /v1/allowance, GET /v1/search/capabilities, POST /v1/search/exposure, and POST /v1/hosts/detail never consume it, regardless of outcome — call them as often as you need.

Within POST /v1/search, a request is charged only when it answers with 200 — including a successful response with zero hosts, because that is an answer too. Every other outcome — a validation error, an auth or plan problem, a rate or capacity limit, a timeout, or an expired cursor — costs nothing and can be retried immediately without spending your allowance.

Read the allowance object on every 200 response instead of polling GET /v1/allowance between searches; it already reflects the search you just ran. Poll GET /v1/allowance only when you have not searched recently and want to check before you do — asking never spends one.

A separate, independent limit caps how many searches your account can have running at the same time; a request beyond it answers 429 user_concurrency_reached and, like every non-200 outcome, is free to retry once an earlier request finishes.

Pagination

POST /v1/search returns next_cursor when more results exist: an opaque sc1_ handle, 36 characters in total. Send it back as cursor in the next request with the same query and limit to get the following page. It is:

  • bound to that exact query and limit — changing either while paging answers 422 validation_failed;
  • one-time — it cannot be replayed once used;
  • short-lived — it expires after a bounded window.

A missing, expired, replayed, or otherwise invalid cursor answers 410 cursor_gone. Start over from an unset cursor; the fresh first page is a normal, chargeable search, but the expired attempt itself was free. A page holds at most 100 hosts, ordered ascending by IP address, so paging through the same query in the same order always produces the same sequence of hosts.

POST /v1/hosts/detail paginates a single host's own services with its own, separate continuation string (up to 512 characters, no fixed format), scoped to that one IP and returned only from that endpoint. It is unrelated to the sc1_ search cursor and the two are not interchangeable.

Endpoints

GET /v1/search/capabilities

The current supported-protocol registry — the same list Add filter loads. Call it rather than hardcoding protocol names, and use catalog_version to detect when the list has changed instead of refetching on a timer; the response is not intended for an HTTP cache (Cache-Control: no-store).

curl -s -H "Authorization: Bearer ipk_live_7pQ2m9XvB4hK1sR6tY0wZ3nC8dF5gJaL" \
     https://api.ip.alphastrike.io/v1/search/capabilities
Field Type Notes
request_id string
query_registry_version string Currently the fixed value service-v5.
catalog_version string Opaque; changes when the supported set changes.
protocols[] array Exactly 87 entries today, sorted lexicographically.
protocols[].value string Canonical lowercase token, for example https.
protocols[].ports[] integer[] Sorted, unique ports this token's catalog covers.
protocols[].ipv4 / .ipv6 boolean Whether the token applies to that address family.
protocols[].evidence_kind string summary, http, or ssh — which typed evidence shape Selected host shows for it. summary and http evidence can carry a tls block; see TLS certificate data.
{
  "request_id": "4f2b6b7a2b6a4c6a9e6e2b7a2b6a4c6a",
  "query_registry_version": "service-v5",
  "catalog_version": "2026-08-01-v7",
  "protocols": [
    { "value": "http", "ports": [80, 8000, 8080], "ipv4": true, "ipv6": true, "evidence_kind": "http" },
    { "value": "https", "ports": [443, 8443], "ipv4": true, "ipv6": true, "evidence_kind": "http" },
    { "value": "ssh", "ports": [22], "ipv4": true, "ipv6": true, "evidence_kind": "ssh" }
  ]
}

(showing 3 of 87 — the response is not truncated; fetch it to see the current full list.)

GET /v1/allowance

Your remaining daily searches, without spending one. Covered fully in Allowance, revocation, and security:

curl -s -H "Authorization: Bearer ipk_live_7pQ2m9XvB4hK1sR6tY0wZ3nC8dF5gJaL" \
     https://api.ip.alphastrike.io/v1/allowance
{ "request_id": "4f2b6b7a2b6a4c6a9e6e2b7a2b6a4c6a", "allowance": { "limit": 20, "used": 6, "remaining": 14, "resets_at": "2026-08-12T00:00:00Z" } }

If a response has no allowance field at all — here or on POST /v1/search — no daily limit applies to that account. That is not the same as none remaining.

POST /v1/search

The matching hosts for a query, exactly like the search box. limit defaults to 25 and accepts up to 100. See Supported searches for the query grammar.

curl -s -H "Authorization: Bearer ipk_live_7pQ2m9XvB4hK1sR6tY0wZ3nC8dF5gJaL" \
     -H "Content-Type: application/json" \
     -d '{"query":"organization:\"Alpha Strike\" protocol:https","limit":2}' \
     https://api.ip.alphastrike.io/v1/search
Field Type Notes
query string 1–512 characters.
limit integer 1–100, default 25.
cursor string From a previous response's next_cursor; see Pagination.

A query for one exact IP address can be answered from a short-lived cache: a found host for up to 15 minutes, an empty result for up to 5 minutes. It never returns a next_cursor, and it is metered like any other search whether or not the cache answered it.

The response is one entry per unique IP. services.items is the bounded five-row preview described in Matching services; a service carries match_reasons only when your query named a service, port, product, certificate, CPE, CVE, CVSS, or exploit clause. organization_matches, domain_matches, and asn_matches appear only for that respective search type.

{
  "request_id": "4f2b6b7a2b6a4c6a9e6e2b7a2b6a4c6a",
  "results": [
    {
      "ip": "203.0.113.10",
      "location": { "country_code": "DE", "country_name": "Germany", "city": "Frankfurt am Main", "latitude": 50.1109, "longitude": 8.6821, "accuracy_radius_km": 50 },
      "network": { "asn": 64496, "organization": "Alpha Strike Labs GmbH" },
      "whois": { "organization": "Alpha Strike Labs GmbH", "netname": "ALPHASTRIKE-NET", "cidr": "203.0.113.0/24", "registry": "ripe", "country_code": "DE" },
      "reverse_dns": "host-10.example.com",
      "domains": { "items": [
        { "name": "host-10.example.com", "source": "reverse_dns" },
        { "name": "www.example.com", "source": "certificate_subject" }
      ], "has_more": false },
      "services": {
        "items": [
          {
            "port": 443,
            "protocol": "https",
            "observed_at": "2026-08-10T03:14:00Z",
            "product": "F5 NGINX",
            "version": "1.26.1",
            "certificate_names": ["example.com"],
            "vulnerabilities": { "cve_count": 0, "has_exploits": false },
            "evidence": [
              { "kind": "http", "status_code": 200, "title": "Example", "server": "nginx",
                "tls": { "version": "TLSv1.2", "cipher_suite": "TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256",
                         "fingerprint_sha256": "b1946ac92492d2347c6235b4d2611184b1946ac92492d2347c6235b4d2611184",
                         "serial_number": "04a1b2c3d4e5f6", "self_signed": false,
                         "subject_common_name": "example.com", "names": ["example.com", "www.example.com"], "names_truncated": false,
                         "issuer_common_name": "Example Fixture CA", "issuer_organization": "Example Trust Services",
                         "valid_from": "2026-06-01T00:00:00Z", "valid_until": "2026-08-30T00:00:00Z",
                         "public_key_algorithm": "ECDSA", "public_key_bits": 256, "signature_algorithm": "ECDSA-SHA384",
                         "browser_trusted": true } }
            ],
            "match_reasons": ["protocol:https"]
          }
        ],
        "has_more": false
      },
      "organization_matches": [
        { "source": "as_organization", "value": "Alpha Strike Labs GmbH" }
      ]
    }
  ],
  "next_cursor": "sc1_ab12CD34ef56GH78ij90KL12mn34OP56",
  "allowance": { "limit": 20, "used": 7, "remaining": 13, "resets_at": "2026-08-12T00:00:00Z" }
}

A domain query adds domain_resolution at the top level instead of organization_matches, and each result carries domain_matches instead:

{
  "domain_resolution": { "status": "resolved", "address_count": 2 },
  "results": [
    { "ip": "203.0.113.10", "services": { "items": [], "has_more": false },
      "domains": { "items": [], "has_more": false },
      "domain_matches": [ { "source": "dns_resolution", "value": "example.com" } ] }
  ]
}

POST /v1/search/exposure

The bounded geographic aggregation behind the Globe view, described in Visualize geographic exposure. Not available for domain queries yet.

curl -s -H "Authorization: Bearer ipk_live_7pQ2m9XvB4hK1sR6tY0wZ3nC8dF5gJaL" \
     -H "Content-Type: application/json" \
     -d '{"query":"organization:\"Alpha Strike\"","resolution":"regional"}' \
     https://api.ip.alphastrike.io/v1/search/exposure
Field Type Notes
query string Any supported non-domain search.
resolution string global (default), regional, or local — coarsest to finest.
{
  "request_id": "4f2b6b7a2b6a4c6a9e6e2b7a2b6a4c6a",
  "unique_hosts": { "value": 842, "relation": "approximate" },
  "service_count": 1310,
  "resolution": "regional",
  "geotile_precision": 6,
  "locations_relation": "complete",
  "locations": [
    { "key": "6/33/22", "latitude": 50.11, "longitude": 8.68, "unique_hosts": 41, "service_count": 63, "country_code": "DE", "country_count": 1 },
    { "key": "6/34/22", "latitude": 48.85, "longitude": 2.35, "unique_hosts": 12, "service_count": 18 }
  ]
}

key is an opaque bucket address (an Elasticsearch geotile coordinate), not a place name — label a cluster from country_code, never from key or from the coordinates. The second bucket above omits country_code and country_count on purpose: that is the documented case where no host in the bucket carries a GeoIP country, and a client must fall back to the coordinates rather than inventing a country. locations_relation: "bounded" means the 500-bucket response limit was reached; unique_hosts and service_count still describe the complete matching result set either way.

POST /v1/hosts/detail

The complete current service inventory for one exact IP — the same data behind Host investigation and Selected host. Never metered, so open it as often as you like.

curl -s -H "Authorization: Bearer ipk_live_7pQ2m9XvB4hK1sR6tY0wZ3nC8dF5gJaL" \
     -H "Content-Type: application/json" \
     -d '{"ip":"203.0.113.10","limit":2}' \
     https://api.ip.alphastrike.io/v1/hosts/detail
Field Type Notes
ip string Exact IPv4 or IPv6 address.
limit integer 1–20 services per page.
cursor string From a previous response's next_cursor; see Pagination.
{
  "request_id": "4f2b6b7a2b6a4c6a9e6e2b7a2b6a4c6a",
  "host": {
    "ip": "203.0.113.10",
    "location": { "country_code": "DE", "country_name": "Germany", "city": "Frankfurt am Main", "latitude": 50.1109, "longitude": 8.6821, "accuracy_radius_km": 50 },
    "network": { "asn": 64496, "organization": "Alpha Strike Labs GmbH" },
    "whois": { "organization": "Alpha Strike Labs GmbH", "netname": "ALPHASTRIKE-NET", "cidr": "203.0.113.0/24", "registry": "ripe", "country_code": "DE" },
    "reverse_dns": "host-10.example.com",
    "domains": { "items": [
      { "name": "host-10.example.com", "source": "reverse_dns" },
      { "name": "www.example.com", "source": "certificate_subject" }
    ], "has_more": false },
    "services": {
      "items": [
        { "port": 22, "protocol": "ssh", "observed_at": "2026-08-10T03:12:00Z",
          "evidence": [ { "kind": "ssh", "protocol_version": "2.0", "software_version": "OpenSSH_9.6", "host_key_algorithm": "ssh-ed25519", "host_key_fingerprint": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855" } ] },
        { "port": 443, "protocol": "https", "observed_at": "2026-08-10T03:14:00Z",
          "product": "F5 NGINX", "version": "1.26.1", "certificate_names": ["example.com"],
          "vulnerabilities": { "cve_count": 0, "has_exploits": false },
          "evidence": [ { "kind": "http", "status_code": 200, "server": "nginx" } ] }
      ],
      "has_more": true
    }
  },
  "next_cursor": "eyJvIjoyLCJpcCI6IjIwMy4wLjExMy4xMCJ9"
}

The answer for an address can come from the same short-lived cache as an exact-IP search: up to 15 minutes for a found host, up to 5 minutes for host_not_found. Pages of one host stay consistent with each other while it is cached.

host_not_found (404) means no current observation exists for that exact address — the same outcome Host investigation shows as "No stored hosts."

TLS certificate data

A service that speaks TLS can carry a tls block inside its evidence. It describes the handshake the scanner completed and the certificate the server presented in it. The block sits on the http evidence of an HTTPS service and on the summary evidence of every other service; there is no separate evidence kind for it. POST /v1/search and POST /v1/hosts/detail return the same block.

{
  "port": 587,
  "protocol": "smtp",
  "observed_at": "2026-11-16T09:00:00Z",
  "certificate_names": ["mail.example.com", "*.example.com"],
  "evidence": [
    {
      "kind": "summary",
      "tls": {
        "version": "TLSv1.2",
        "cipher_suite": "TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256",
        "fingerprint_sha256": "3f2a9c0d1e4b5a69788796a5b4c3d2e1f00112233445566778899aabbccddeef",
        "serial_number": "04a1b2c3d4e5f6",
        "self_signed": false,
        "subject_common_name": "mail.example.com",
        "names": ["mail.example.com", "*.example.com"],
        "names_truncated": false,
        "issuer_common_name": "Example CA R3",
        "issuer_organization": "Example Trust Services",
        "valid_from": "2026-10-01T00:00:00Z",
        "valid_until": "2026-12-30T23:59:59Z",
        "public_key_algorithm": "RSA",
        "public_key_bits": 2048,
        "signature_algorithm": "SHA256-RSA",
        "browser_trusted": true
      }
    }
  ]
}

Every field is optional. A value the scan did not record, or that fails the checks below, is left out; it never arrives as an empty string, an empty list, or a placeholder.

Field Type Notes
version string Negotiated protocol version, for example TLSv1.2. The scanner offers TLS 1.2 at most, so TLSv1.3 never appears, also for servers that support it.
cipher_suite string Negotiated cipher suite, IANA name.
fingerprint_sha256 string SHA-256 over the DER encoding of the server certificate, 64 lowercase hexadecimal characters.
serial_number string Certificate serial number as lowercase hexadecimal with an even number of digits, no prefix and no separators, at most 64 digits. A negative serial number, which the standard does not allow, is left out.
self_signed boolean true when the certificate's signature verifies with its own public key. This is the scanner's signature check, not a comparison of subject and issuer names.
subject_common_name string Common name of the certificate subject, at most 256 characters.
names string array Names the scanner read from the certificate: the entries of its subject alternative name extension, plus the subject common name when it looks like a host name. Lowercase, no duplicates, wildcard entries unchanged, at most 32 entries. It is not a list of alternative names only; the scan data does not keep those apart.
names_truncated boolean Present whenever names is. true when the certificate holds more names than names returns.
issuer_common_name string Common name of the issuer, at most 256 characters.
issuer_organization string Organization of the issuer, at most 256 characters.
valid_from, valid_until string Validity period, RFC 3339 in UTC.
public_key_algorithm string RSA, ECDSA, or DSA.
public_key_bits integer Key size in bits for RSA and ECDSA keys. Not recorded for DSA.
signature_algorithm string One of MD2-RSA, MD5-RSA, SHA1-RSA, SHA256-RSA, SHA384-RSA, SHA512-RSA, DSA-SHA1, ECDSA-SHA1, ECDSA-SHA256, ECDSA-SHA384, ECDSA-SHA512. Any other algorithm is left out.
browser_trusted boolean true when the scanner could build a chain from the certificate to a root in its browser root store. The host name is not compared, and the result reflects the time of the scan.
browser_untrusted_reason string Only when browser_trusted is false: unknown_authority, expired_or_not_yet_valid, or not_authorized_to_sign. It is the first problem the validation reported; there may be others, and it is left out when the scanner gave no specific reason. A self-signed certificate shows self_signed: true together with unknown_authority.

What to know when you build on the block:

  • Which services carry it. HTTPS and the other web protocols observed over TLS, SMTP, IMAP and POP3 (on the directly encrypted ports and after STARTTLS), MySQL, Microsoft SQL Server, PostgreSQL, RDP, OpenVPN, SoftEther VPN, and Cobalt Strike. PostgreSQL and RDP services carry the certificate fields only, without version, cipher_suite, and the two trust fields. No certificate is recorded for LDAP over TLS, FTP over TLS, AMQP over TLS, WinRM over TLS, or SIP over TLS, so those services have no tls block.
  • Which certificate it is. Always the server's own certificate, never one from the chain it sent, and always from the same observation as the service's observed_at.
  • No server name. The scan addresses the IP address and records no server name for the handshake. On an address that serves several names, the block shows the certificate the server presents by default.
  • Completed observations only. A service whose scan ended with an error has no tls block, because the handshake may be incomplete. Its certificate_names can still be present.
  • certificate_names is a different list. It holds at most eight names per service, without a truncation flag. Use tls.names with names_truncated when you need to know whether the list is complete. Both lists hold plain ASCII names in lowercase.
  • Text from certificates. A certificate is written by whoever runs the server. A subject, issuer, or name that contains control characters or invisible formatting characters is left out instead of being repaired. A name that is not plain ASCII is left out of names as well, and names_truncated is then true; internationalized host names appear in their xn-- form.
  • New fields. The block gained most of its fields in API version 0.28.0 and may gain more. Read the fields you know and ignore the others. A client that validates responses strictly against an older schema has to update that schema first.

Errors

Every error uses the same shape:

{ "request_id": "4f2b6b7a2b6a4c6a9e6e2b7a2b6a4c6a", "code": "validation_failed", "message": "request validation failed", "violations": [ { "field": "query", "code": "invalid_organization" } ] }

violations is only ever non-empty for validation_failed, and it is how you find out which part of the request was rejected — code alone tells you the request was invalid, not why.

Top-level codes

code HTTP status Meaning Costs your allowance?
invalid_json 400 The body is not valid JSON. No
organization_required 400 Your account does not currently resolve to exactly one organization. Contact us. No
unauthorized 401 The key is missing, malformed, unknown, or revoked. No
plan_expired 402 Your plan does not currently include search access. No
validation_failed 422 The query, limit, resolution, or cursor is invalid — see violations. No
search_limit_reached 429 Your daily allowance for the current period is used up. No
user_concurrency_reached 429 You already have the maximum number of searches running at once. Wait for one to finish. No
cursor_gone 410 The search cursor is missing, expired, already used, or invalid after a restart. No
host_not_found 404 No current host exists for this exact IP (/v1/hosts/detail only). — (this route is never metered)
search_capacity_exhausted 503 The service is at capacity and declined the search before starting it. No
commercial_access_unavailable 503 Your account's plan could not be confirmed. This is on our side. No
search_source_busy 503 The stored scan data was temporarily busy. Wait a few seconds and retry. No
service_unavailable 503 A dependency the request needs is unavailable. No
audit_unavailable 503 The mandatory audit record could not be written, so the request was not run. No
search_timeout 504 The search did not finish within its time budget. Narrow the query or retry. No

As Rate limits and the daily allowance explains, the only outcome that ever costs you a search is a 200 from POST /v1/search — every code in this table is free.

Violation codes (validation_failed)

These appear in violations[].code with violations[].field set to "query" unless noted.

code Meaning
invalid_query_syntax The query does not parse as a supported expression.
unknown_field A field name is not one of the approved filters.
duplicate_field The same field appears more than once in one query.
unsupported_operator Something other than a space or uppercase AND was used to combine clauses.
invalid_port The port is missing, non-numeric, or out of range.
unsupported_protocol Not an exact value from GET /v1/search/capabilities.
invalid_organization The organization phrase is missing its quotes or is otherwise malformed.
invalid_product The product phrase is missing its quotes or is otherwise malformed.
invalid_certificate_name Not one exact DNS name, or a wildcard was used.
invalid_country Not an uppercase ISO-2 code.
invalid_cpe Not one complete, quoted CPE 2.3 URI.
invalid_cve Not the exact CVE-YYYY-NNNN form.
invalid_cvss Not an inclusive min-max range within 0.0–10.0.
invalid_exploit Not exactly yes or no.
invalid_cidr The network does not parse as an IPv4 CIDR.
cidr_too_broad The prefix is broader than /24.
ipv6_cidr_not_supported IPv6 CIDR search is not available yet.
invalid_domain Not a registrable domain name, or a wildcard or public-suffix-only value was used.
invalid_asn Not the standalone decimal form asn:<1..4294967295>.
query_too_broad The query names no service and matches more hosts than one search can return (POST /v1/search only — combine it with a port, protocol, or country).
unsupported_query The query does not match any supported search type.
exposure_not_supported This query type cannot be aggregated yet (POST /v1/search/exposure only — domain queries are excluded).
unsupported_resolution (field: "resolution") Not global, regional, or local (POST /v1/search/exposure only).
invalid_ip (field: "ip") Not a valid IPv4 or IPv6 address (POST /v1/hosts/detail only).
invalid / query_mismatch (field: "cursor") The cursor is malformed, or does not match the query and limit it was issued for (POST /v1/search only — a cursor problem otherwise answers 410 cursor_gone, not this).

The Troubleshooting searches table describes the same conditions in the wording the web app shows for each one.

Building a client against this API

  • Handle every code in the tables above, not just the ones you expect during development — service_unavailable, search_source_busy, and search_timeout are ordinary operating conditions for a bounded, best-effort search service, not edge cases.
  • Only plan_expired, organization_required, and unauthorized need a human to act. Every other non-200 response — timeouts, capacity, concurrency, the daily limit, and an expired cursor — is safe to retry: back off briefly for a 503 or 504, and wait for resets_at before retrying search_limit_reached.
  • Page with next_cursor until it is absent, and treat 410 cursor_gone as "start again from an unset cursor" rather than an error to surface to a person.
  • Read allowance from each response rather than tracking your own count; it is authoritative and already includes the request that returned it.
  • Include request_id in anything you log or report — it is the fastest way to have a specific request investigated.

A minimal pagination loop, stopping when the allowance runs out or the query is exhausted:

import requests

API = "https://api.ip.alphastrike.io"
KEY = "ipk_live_7pQ2m9XvB4hK1sR6tY0wZ3nC8dF5gJaL"

def search_all(query, limit=100):
    cursor = None
    while True:
        body = {"query": query, "limit": limit}
        if cursor:
            body["cursor"] = cursor
        response = requests.post(
            f"{API}/v1/search",
            headers={"Authorization": f"Bearer {KEY}"},
            json=body,
            timeout=30,
        )
        if response.status_code != 200:
            error = response.json()
            raise RuntimeError(f"{error['code']}: {error['message']} (request_id={error['request_id']})")

        payload = response.json()
        yield from payload["results"]

        allowance = payload.get("allowance")
        if allowance and allowance["remaining"] <= 0:
            break
        cursor = payload.get("next_cursor")
        if not cursor:
            break

for host in search_all('organization:"Alpha Strike" protocol:https'):
    print(host["ip"])

This omits retry/backoff for brevity; add it around the requests.post call for 503 and 504 per the guidance above before relying on this in production.