Skip to content
New Pro plans with Agentic Verification.

Agentic Verification. Proof your software builds. In hours, not weeks.

AI agents bring your escrow deposit into a sealed lab, figure out how it's built, and then build it, with a dated Recoverability Certificate issued for each passed run.

Part of Escrow Pro, Resilience Pro, Continuity Pro, and Software Backup Pro, with as many as four runs a year on supported technology stacks.

Book a demo
  • Runs on Codekeeper's own AI, apart from public models
  • No internet connection
An engineer watches a rebuild finish in the verification lab

Payments Core · Run #11

Escrow Pro · Codekeeper AI Verification Lab

  1. Pre-flight checkStack matched
  2. Seal the environmentSealed
  3. Restore the depositRestored
  4. Work out the build1 private package resolved
  5. Isolated compileCompiled from source
  6. Start and checkHealth checks passed
  7. Write the evidence4 documents

Passed · Sep 3, 2026

Recoverability Certificate issued

A sample Agentic Verification run for an application named Payments Core, using example values, covering seven steps from the pre-flight check to the evidence, passed on 3 September 2026 with a Recoverability Certificate issued.

Relied on by 3 500+ teams across more than 150 countries, from regulated enterprises to software vendors growing at pace

Airbus Bayer European Parliament General Motors Intuit Nestlé PepsiCo Pfizer

What's new

Rebuild tests used to be a project. Now they're a button.

Testing a build by hand takes weeks of specialist effort, engineers to schedule, and quotes to sign off, which is why it was priced for the biggest contracts. Agentic Verification takes hours, starts when you press run, and comes with every Pro plan, as many as four times a year.

Compared with a typical manual build test

Before

A typical manual build test

Specialists, booked per engagement

  • TimeWeeks of specialist work
  • Starting a testEngineers to book, quotes to approve
  • CostPriced for the largest contracts
  • The vendorOften relies on the vendor's engineers
  • EvidenceA report at the end of the engagement

Now

Agentic Verification

AI agents, in the Codekeeper lab

Included in Pro
  • TimeHours
  • Starting a testPress run, whenever you choose
  • CostUp to four runs a year included, on supported technology stacks
  • The vendorNo vendor help on a complete deposit
  • EvidenceA report every run, and a certificate for every passed run

For specialists who recover the software and test your chosen functions, see the Exit Exercise.

Inside a run

Press run. The agents do the rest.

Your deposit becomes a certificate with no one to schedule, for as many as four runs a year on supported technology stacks, each launched when it suits you.

  1. 1

    Pre-flight check

    The agents read the deposit and match each part to a supported technology stack.

  2. 2

    Seal the environment

    A clean environment opens in the Lab, for this run only.

  3. 3

    Restore the deposit

    Repositories and documents come back as deposited.

  4. 4

    Work out the build

    The agents study how it's built and resolve its dependencies, private packages included.

  5. 5

    Isolated compile

    The software is compiled from source, inside the sealed environment.

  6. 6

    Start and check

    Services start, and health checks confirm the application runs.

  7. Write the evidence

    Rebuild Report, SBOM, and exit workbook. A pass adds the Recoverability Certificate.

A typical run, illustrated; the precise steps vary with the technology stack.

Unsure whether your stack is supported? Take the free technical assessment, and once you've told us how your application is built, our escrow engineers will advise what to deposit and which verification level suits you.

0:00 / 1:11

Starting a run

Launch it when it suits you.

  • Run it at the moment you pick, such as ahead of an audit, after a release, or once the deposit changes.
  • No engineers to schedule and no quotes to sign off.
  • Between runs, we check the deposit against the live repositories and point out what's changed, so you know when to run it again.

What you get

Four documents. One answer you can hand over.

Each passed run adds a dated set to your records, with every document answering a question you'll be asked.

The sample evidence pack for an example run: a Recoverability Certificate, a Rebuild Report, and SBOM and exit workbook excerpts
Shown with example values for an invented application: a Recoverability Certificate and Rebuild Report. For tests of your everyday functions, see the Exit Exercise.

  • Recoverability Certificate

    Did it build, and when?

    The application, the result, and the date, on one page you can hand to an auditor or a client.

  • Rebuild Report

    How did it build?

    Each step the agents took, what compiled, and anything that was missing.

  • SBOM

    What's inside it?

    Every component the build used, direct and transitive, with versions and licenses, in CycloneDX.

    More about SBOMs →

  • Exit workbook

    How would we recover it ourselves?

    The recovery, step by step, written from the run: environment, deposit, build, configuration, and checks.

  • Plus a tested isolated compile

    Is it ready to deploy?

    Every passed run also produces a ready-to-deploy build. It's held with the deposit until a release, when the software user gets it.

Look at the evidence before speaking to us.

The sample evidence pack contains the Recoverability Certificate and Rebuild Report from a single example run, plus SBOM and exit workbook excerpts, all using example values.

Who it's for

Proof for both sides of the agreement.

The software user and the vendor get the same dated evidence from the same run.

An IT risk lead walks an external auditor through a printed report at a meeting table

01 · Software users

You depend on the software.

Questions from auditors, risk teams, and your board.

  • Confidence that the software you rely on can be rebuilt
  • Dated evidence for audit and risk reviews
  • Gaps caught while the vendor can still fill them
See the plans that include it →
A software vendor's engineering lead answers a client's security review on a video call, with the questionnaire on a second screen

02 · Software vendors

You build the software.

Asked in client security reviews and procurement.

  • Answer client security reviews with a certificate
  • Show what's in your deposit before a client asks
  • Get an exit workbook that shows clients their recovery is already planned
Codekeeper for software vendors →

When a run finds a gap

Found a gap? You found it early.

The Rebuild Report identifies it, the vendor puts it in the deposit, and you start another run.

Run #10Payments Core1 item flagged
  • Pre-flight check
  • Deposit restored
  • Dependenciesledger-core 4.2, a private package, isn't in the deposit
Flagged in the Rebuild Report for the vendor
  1. 1 SepVendor informed
  2. 2 SepPackage added to the deposit
Run #11Payments Core · Sep 3, 2026Passed
  • Dependencies resolvedIncluding 1 private package
  • Isolated compile
  • Started, health checks passed
Recoverability Certificate issued

Example values, matching those in the sample evidence pack.

Security

Sealed. Private. Our own AI.

Your code never leaves Codekeeper's environment. Every run is sealed and uses Codekeeper's own AI, which keeps your code away from public models.

Codekeeper AI Verification Lab

A sealed environment for each run. Inside it:

  • Your deposit
  • Codekeeper's AI agents
  • Codekeeper's own models

No connection to

  • Production systems
  • Development systems
  • The open internet
  • Public AI models

Our models, our infrastructure.

Codekeeper's environment hosts the agents and the models they run on.

The vendor keeps control.

Materials can be added or removed by the vendor at any time, and each run works only from what's present.

Certified independently.

Codekeeper holds ISO/IEC 27001:2022 certification.

Plans

Part of every Pro plan.

Escrow Pro, Resilience Pro, and Continuity Pro each come with as many as four runs a year on supported technology stacks, plus the complete set of evidence, so pick the one that matches how you'd like to recover. Software Backup Pro includes it as well, for software your own team builds and runs.

These plans include Agentic Verification. From prices per application type: Software, SaaS, and AI.
PlanSoftwarefromSaaSfromAIfrom
Escrow ProSelf-Run RecoveryYour team receives the released deposit and carries out the recovery.$419$599$1,019
Resilience ProManaged RecoveryThe deposit is released to Codekeeper, whose engineers handle the recovery.$599$789$1,299
Continuity ProFull ContinuityWith Escrow Pro or Resilience Pro, we pay for the live environment and keep it running for an agreed period, recharging the bills at cost.$1,569$1,799$2,219
Software Backup ProYour own softwareFor applications your own team builds and runs, held in an immutable vault beyond production.$319billed annuallyNot offeredNot offered

One-time setup: $499 per agreement for Escrow Pro and Resilience Pro, $999 for Continuity Pro, and $249 for Software Backup Pro.

Base plans (Escrow, Resilience, and Continuity) carry a monthly automated check with its Vault Report and Vault Certificate, but rebuilds come only with Pro.

Compare plansBuild a plan →

Have the answer ready before anyone asks.

Try it in a demo on one of your own applications, or open the sample evidence pack first.

Book a demo

Frequently asked questions

What to know ahead of your first run.

How does Agentic Verification work?

Agentic Verification works by having Codekeeper's AI agents rebuild your deposit in the Codekeeper AI Verification Lab and record the outcome, as the rebuild test in every Pro plan. A passed run issues a Recoverability Certificate.

Could a public AI model ever see our code?

No, a public AI model never sees your code. The agents work on Codekeeper's own AI and infrastructure, keeping your code inside our environment the whole time.

Is the vendor's help needed?

The vendor's help isn't needed when the deposit is complete. Agents rely on what's in the deposit, and any missing item gets flagged for the vendor to add.

What technology stacks can you rebuild?

The technology stacks we can rebuild are confirmed for your application in the demo, so let us know what it's built with. Every run's pre-flight check also pairs each part of the deposit with a supported stack before building begins.

How many runs do we get?

You get as many as four Agentic Verification runs a year on each Pro plan, on supported technology stacks. The timing is yours, so a run can come just before an audit or straight after a major release.

Does a Recoverability Certificate prove compliance?

No, a Recoverability Certificate doesn't prove compliance. It records that the deposit built and the date it did, and you can include it in the evidence for an audit or review, though you and your assessor decide whether it meets a particular requirement.

What happens when a run doesn't pass?

When a run doesn't pass, the Rebuild Report shows where the build stopped and which parts were missing. Once the vendor supplies them, start a fresh run.

What separates this from an Exit Exercise?

What separates Agentic Verification from an Exit Exercise is the question each answers. Agentic Verification shows whether the deposit builds, while an Exit Exercise shows whether you could exit and carry on working, with Codekeeper specialists rebuilding the software independently of the vendor, restoring the data, deploying it, and testing the business functions you choose. An Exit Exercise can be added to any plan.