Platform

The system of record for storage access

One telemetry layer: S3 server access logs, S3 Inventory, and network metadata. reCost turns them into a queryable record of every request, attributed to an identity and a client, retained for years.

reCost access record

What the engine produces

Identity attribution

Every request resolved to its principal: role, assumed-role session, IAM user, external account, or anonymous. Credential type classified, long-lived keys separated from temporary STS sessions.

reCost identities, resolved to principal and credential type
reCost client fingerprint changes, by identity

SDK and runtime fingerprints, with support status and CVE severity per identity.

Client fingerprinting

Every request carries a user agent. reCost parses it to the SDK, runtime, version, and operating system, then tracks that fingerprint per identity over time. End-of-life versions are flagged against vendor support dates, known CVEs matched to detected versions with severity, and a fingerprint change on an established identity surfaces as its own signal.

Access classification

Every identity and user agent pair assigned an access class: application, data platform, ETL, backup, CDN, IaC, human console, AI agent, external crawler, and more.

reCost access classes, by volume
reCost access by class, 30-day trend

Behavioral baselines

Volume, operation mix, timing, and asset set learned per identity. This is what the detections measure against.

Network context

Source IP, subnet, and region for every request, with internal and external traffic separated.

reCost external access, by bucket and client
reCost lineage view, every touch on an object over time

Object lineage

Supply a key, a date, an IAM identity, a role, an agent, or a service, and receive every touch on it, going back as far as your retention runs.

What reCost deliberately does not do

No blocking

reCost observes. Your enforcement stack acts on what it finds.

No content reads

Metadata only. Objects are never opened, never sampled, never classified.

No sampling

Every request, every object. Not a percentage, not a scan window.

No agents

Nothing deployed in your environment. A read-only role and nothing else.

The tributary,
not the river.

Storage access logs are too voluminous to put in a SIEM. reCost processes them outside it, at object level, and sends only what matters downstream.

Your S3 access logs
Billions of requests per month, staying in your account
reCost engine
Classification, baselining, detection, record
Your SIEM / SOC
Curated findings only

Start with proof,
not a pitch.

Scoped read-only role, 30-day lookback, results in 48 hours.