Security & Trust

WhyStack investigates your production infrastructure with an AI agent, so it is built around a single principle: your data and your credentials stay in your environment. Here is exactly how.

Your credentials never leave your infrastructure

The WhyStack Agent runs inside your own cloud. It reads secrets from your secret manager via workload identity and holds them only in memory. The control plane never stores, sees, or custodies a customer credential. No static keys baked into images.

PII is redacted at the edge

Production data is redacted at the Agent boundary, the single enforcement point, before anything crosses your network perimeter. Every redaction is audit-logged. What leaves your environment is redacted metadata, not raw production data.

Read-only by default

AWS, GCP, Datadog, GitHub and every observability integration are strictly read-only, enforced at the connector layer. The agent investigates, but it cannot modify, delete, or reconfigure your infrastructure.

Writes are records, never infrastructure

The only write actions open a Jira issue, post to Slack, or open a GitHub issue: records in tools your team already uses, never a change to your infrastructure. Every write an investigation initiates requires explicit confirmation: the agent shows a preview, a human approves. All of them are recorded in an audit trail. Bounded autonomy, always.

Verifiable identity, everywhere

Agents enroll via workload identity with mutual TLS against a private CA, with no shared secrets. People authenticate with role-based access control and 2FA. The agent acts as a scoped user; write permissions are fail-closed (denied unless explicitly granted).

Right to erasure

Anything WhyStack learns is identity-bound, decays over time, and can be erased on request. Deleting your organization cascades a full purge of your data across every store.

Where your data lives

WhyStack runs in a split architecture. The part that touches your infrastructure lives in your cloud; the part that coordinates lives in ours. The boundary between them is the whole point.

Runs in your cloud

The Agent is an ephemeral container in your infrastructure. It talks only outbound, over an authenticated mutual-TLS tunnel, so you never open an inbound port to us.

Redacted at the boundary

Before any result leaves your network, PII is redacted at the Agent. The redaction happens on your side of the line, not ours.

We hold metadata, not secrets

The control plane stores redacted investigation metadata and audit logs. Cross-customer intelligence is anonymized and strictly opt-in, and off by default.

Compliance & posture

We are transparent about what is in place today and what is on the way.

SOC 2

Readiness assessment complete against the AICPA Trust Services Criteria; Type I targeted for launch, Type II to follow. Much of the control set is architectural: credentials never custodied, PII redacted at the edge, audit logging, RBAC, erasure.

Encryption in transit

Mutual TLS between the Agent and the control plane; TLS for every external surface. No plaintext across trust boundaries.

Auditability

Every write action and every PII redaction is logged with the acting identity. Investigations are traceable end-to-end.

Privacy by default

Data minimization at the edge, identity-bound retention with automatic decay, and opt-in-only cross-customer learning. The private option is always the default.

Security questions?

We are happy to walk your security team through the architecture, share our readiness documentation, or scope a review.