k8sKubernetes
Every create, update and delete at admission, on any conformant distribution.
Change evidence for everything you run
Stonewake records each change to your clusters, cloud accounts, pipelines and databases at the moment it is made, checks it against your policies, and seals the result into a tamper-evident evidence pack. When an auditor samples a change, you send a link instead of losing a week.
sha256:…
chain verified
The problem
Auditors do not test your dashboard. They pick individual changes and ask what happened to each one. Today someone on your team rebuilds that answer by hand, for every system the change touched.
Reporting clocks now run in hours. Evidence that takes days to assemble arrives too late.
Live demo · runs in your browser
Pick a change and apply it. The digests below are real SHA-256 hashes computed on this page, and each pack is chained to the one before it. Then try to tamper with one.
Coverage
A production change rarely stays in one system. The same release touches a pipeline, a cluster, a cloud account and a database. Stonewake gates each of them and links the records into one story.
k8sEvery create, update and delete at admission, on any conformant distribution.
cloudPlans and applies for AWS, Azure and Google Cloud, plus console changes made outside code.
ci/cdMerges, deployments and protected-branch overrides in GitHub, GitLab and Argo CD.
dataSchema migrations and privileged statements on PostgreSQL and MySQL.
identityRole grants, policy edits and admin actions in Okta, Entra ID and your Git organisation.
agentsTool calls that change production, recorded with the agent, the model run and the human behind it.
The platform
Pipelines, infrastructure-as-code and AI agents now make most production changes. Stonewake treats each of them as an accountable actor and keeps proof you can hand to someone outside your team.
Every pack carries the digest of the pack before it. Change one field in one record and every later digest stops matching. Ledger heads are signed on a schedule and can be anchored to storage you control, such as an object-lock bucket.
append-only · hash-chained · signed heads
Risky change? The approver gets it in Slack, Teams or ServiceNow with the diff and the finding. The signed token covers that one digest and expires in four hours. Change one line and it needs a new approval.
Ed25519 · digest-bound · expiring
When an agent or a pipeline makes the change, the pack names the workload identity, the human who delegated it and the run that triggered it. "The bot did it" stops being the end of the trail.
SPIFFE · OIDC · delegation chain
Exceptions need an owner, a scope, a reason and an end date of 90 days or less. Owners are reminded before expiry. Nothing stays waived forever because someone forgot.
scoped · justified · 90-day max
Collectors record what a workload does after the change lands: processes and connections on nodes and hosts, drift on cloud resources. A shell in production or an unexpected outbound connection is attached to the change that shipped it.
eBPF · drift detection · no sidecars
Give your auditor a read-only, time-boxed view of the sampled population. They pick the changes, open each pack and verify the digests themselves. No screenshots, no spreadsheet exports.
read-only · time-boxed · self-verify
Audit records outlive the cryptography that signed them. Packs can carry a hybrid signature that pairs Ed25519 with ML-DSA (FIPS 204), so evidence sealed today still verifies after classical signatures are retired.
Ed25519 + ML-DSA-65 · crypto-agile
| What the auditor needs | Manual assembly | Policy engine alone | Stonewake |
|---|---|---|---|
| Record of every change, not only violations | Partial, from logs | No | Yes |
| Proof the record was not altered | No | No | Hash chain + signature |
| Approval tied to the exact object | Ticket number only | No | Digest-bound token |
| Agent and pipeline identity | Service account name | Service account name | Full delegation chain |
| Mapped to control IDs | By hand, per audit | No | In every pack |
| Time to answer one sample | Hours | Hours | Seconds |
Rollout
Every change from every connected source is captured and packed. Nothing is blocked and nobody's deploy slows down. You see what your policies would have caught.
Engineers get a warning on risky changes at apply time, with the reason and the fix. Outcomes are recorded. Still no blocking.
Policy by policy, changes above your threshold need an approval token or an active waiver. Everything else passes untouched.
One command per source. Every gate fails open by default, so a Stonewake outage never stops a deploy.
stonewake connect kubernetes --context prod-east-2 --mode record stonewake connect aws --org o-example --mode record stonewake connect github --org your-org --mode record stonewake connect postgres --host ledger-db.internal --mode record
Frameworks
Choose your framework to see which controls a pack answers and what the auditor receives.
What it is worth
Move the sliders to match your audits. The estimate assumes about two minutes of review per sampled change once packs exist.
An estimate from your inputs, not a guarantee.
Pricing
Priced per connected source: a cluster, a cloud account, a pipeline organisation or a database. No charge per node, per record or per pack.
$0
For trying it on a real system.
$1,200per source / month, billed yearly
For passing an audit this year.
Custom
For regulated and air-gapped estates.
The pilot promise. Bring one evidence request from your last audit. If Stonewake cannot answer it from a live system within 30 days, you pay nothing and keep the Community tier.
Our own security
The control plane, the ledger and the collectors all run in your environment. There is no Stonewake cloud to breach and no telemetry unless you switch it on.
Every release ships with a signed image, an SBOM and build provenance. Verify them before anything touches your environment.
Connectors start read-only and see change requests and nothing else. Enforcement permissions are added per source, only when you enable enforcement.
No trackers, no cookies and no third-party scripts or fonts. A strict content security policy blocks anything not served from this origin. Open your browser's network tab and check.
Found a vulnerability? See security.txt for how to report it.
Questions
Not unless you ask it to. Record mode adds a check measured in milliseconds and never rejects a request. Every gate fails open, so if Stonewake is down your deploys continue and the gap itself is recorded.
No. Keep your policy engine. Stonewake records its verdicts alongside its own and turns them into evidence. Policy engines decide; they do not keep a per-change record an auditor can verify.
A pack per sampled change as PDF and JSON, or a read-only Auditor Room link. Each pack states what changed, who requested it, which policies ran, who approved it, which waivers applied and which control IDs it supports. Open the sample.
Kubernetes 1.27 or later on any conformant distribution, AWS, Azure and Google Cloud, Terraform and OpenTofu, GitHub, GitLab and Argo CD, PostgreSQL and MySQL, Okta and Entra ID, and AI agents through a tool-call gateway. Air-gapped installs are supported on the Enterprise plan.
The first source is typically recording within an afternoon. Most teams stay in Record mode for a week or two, then turn on advisories, then enforce a handful of policies.
They can change bytes, but they cannot hide it. Every pack includes the previous pack's digest and ledger heads are signed, so any edit breaks verification for every later record. The demo above shows exactly that.
Get started
In 30 minutes we connect Stonewake to one of your systems with you, ship a change and answer the request live. If it does not fit, you will know before the call ends.