Skip to content

Cookie preferences

Choose which cookies and similar technologies this site may use. The strictly necessary ones keep the site working and remember this choice, so they cannot be turned off; they store nothing that identifies you or follows your browsing. Read the cookie notice

Continuumship
Menu

Release and evidence

Every release watched, and proven.

Guards measure a release at the source while it runs, and the response is approved with the release. The evidence chain shows every link from the intent as accepted to the outcome as judged.

Talk to us Features
Guards measured at the source through the window, and the response approved with the release.

Why it matters

An audit should not be an excavation.

  • Nobody watches a release once the pipeline is green.

    Errors climb for twenty minutes before someone notices, and the rollback is decided in a hurry.

  • Our evidence is spread across six tools.

    The auditor asks who approved what, on which evidence. The answer is a week of screenshots.

  • Our cluster runs whatever the pipeline pushes.

    Nothing proves that the image being admitted passed what the release required.

How it works

Watch the release, keep the chain, sign the result.

Guards and response clauses

A guard is a measurement with a bound, evaluated at its source during an observation window. The response — notify, hold or roll back — is approved with the release.

  • The guard plan is part of what is approved
  • A guard that cannot be measured waits on a person
  • When a guard fails, what changed in the window, in order

The evidence chain

Every release carries its chain, from the intent as accepted to the outcome as judged. Each link is present, not required, missing, unknown or broken, and says which.

  • Every claim with its provenance
  • Evidence exported with hashes
  • Compliance and reconstruction packs an auditor can take

Attestations and data placement

The platform signs a statement that an artifact passed what was required, and your cluster's admission rule can require it. Each class of data is placed where you decide.

  • Signed attestations bound to the artifact
  • One page says where each class of data is
  • Evidence kept on your side, with the cost shown first

Try it

Break a release and watch the guard respond.

Push the errors past the bound at any minute of the rollout, choose the response approved with the release, and see what the guard does and what the evidence chain records.

Break the release
+12 per minute
Minute 10
9 min
Response if the guard trips

Approved with the release: roll back to zero when errors stay over 12 a minute for 5 minutes.

Guard tripped

Tripped at minute 14. Rolled back to zero.

Over the bound from minute 10 to 14 · rollback recorded in the evidence chain

Server errors per minute Measured at the source
  1. 1 in 100
  2. 1 in 20
  3. 1 in 4
  4. Everyone

Questions about release and evidence

What can guards measure?
Guards read at the source through your connections: the platform's own health checks, Sentry, Grafana and New Relic. Each guard has a bound and a window, approved with the release.

Can a release roll back on its own?
Only when its response clause says so and a person approved that clause with the release. Rollback applies to managed deployments; otherwise the response is to notify or to hold.

What does an auditor receive?
The evidence chain for each release, with every link's state and provenance, exported with hashes, and a reconstruction pack that rebuilds a work item's lifecycle from its record.

Where is our evidence kept, and who signs the attestation?
Each class of data has a placement row saying where it is and which copy is the authority. Evidence can be kept on your side. The platform signs attestations; your cluster's own rule decides what to admit.

Keep reading

The rest of the control plane.

Design how your company ships software.

Start with a thirty-day shadow pilot on one Area: connect, observe, read the report together, decide.

Talk to us Pricing