Fix the backlog. Prove every step.

Akesis reads the findings your scanners raise, re-scores them for your code and turns the fixes into pull requests. It merges on its own only as far as you allow, and records every decision it makes.

Now onboarding design partners.

θ 000°MEND · 3D
Drag to rotate · ClickTap to re-mend
seed —

One backlog,
zero noise.

Every scanner, one backlog

Reports of the same flaw (same package, same CVE) merge into one finding with every source kept, then leave as a fix on GitHub, GitLab or Bitbucket. Dependabot, Snyk, Strix and any scanner that writes SARIF today; Wiz and Inspector next.

Re-scored for your code

CVSS is where a finding starts. Exploit likelihood (EPSS), CISA’s known-exploited list, whether your code imports the vulnerable package, and exploits confirmed against your own application move it up or down, with the reasons written out.

Fixes arrive as pull requests

Akesis takes the fix PRs Dependabot, Snyk and Renovate open, and opens its own when none exists: the smallest version change, with the lockfile rebuilt in your CI. Every fix is checked against the package’s published checksum. Passing tests are never presented as proof.

Watch the backlog shrink

The backlog shows what was fixed in the last 30 days, how much Akesis fixed on its own and the median time to fix. Progress you can show, not claim.

An audit trail nobody can edit

Every decision, merge, sign-in and policy change is chained by hash. Export it and verify it without us.

Works with

Scanners
Dependabot, Snyk and Strix (AI penetration testing), plus any scanner that writes SARIF: Trivy, Grype, OSV-Scanner, CodeQL, Semgrep and Checkov.
Code hosts
GitHub, GitLab and Bitbucket, with fix pull requests opened, checked and merged where your policy allows.
Tickets and alerts
Jira issues for findings that can’t become a pull request, closed when the scanner confirms the fix. Slack alerts and a daily digest.
Fix data
For any CVE, the vendor’s own fix: Microsoft KB updates, Red Hat, Ubuntu and Debian package versions, PAN-OS releases, Cisco security advisories and CISA’s required action.
AI assistants
Ask Claude about your backlog through Akesis’s read-only MCP server, signed in with your Akesis account.

Autonomy you grant one step at a time.

Each repository has its own ceiling. Raising it needs a reason and is recorded; lowering it never asks.

  1. T0

    Suggest

    Explains the fix on the pull request. Changes nothing.

    Where every new repository starts.

  2. T1

    Draft PR

    Opens a draft for a person to finish.

    Available without a test suite.

  3. T2

    Open PR

    Opens a pull request ready for review. A person merges.

    A test suite Akesis can run against.

  4. T3

    Auto-merge

    Merges when every gate passes: green CI, verified provenance, inside the paths you allowed.

    A verified test suite, a rule you wrote and a stated reason.

What it will never do on its own.

Emergency stop
Any engineer can stop every repository at once. Akesis drops to suggestions immediately, including merges already approved.
Model-written code
Never auto-merged. A person reviews every change a model wrote, whatever the repository allows.
Merges
Re-checked at the moment of merging and pinned to the exact commit CI tested.
Regressions
A failing default branch after an automatic merge stops autonomy for that repository and offers a one-click revert.
Escalations
Raising a tier needs a stated reason and is recorded. Lowering one never asks.
Lowering a priority
Only Akesis’s own analysis of your lockfile can mark a finding unreachable. A scanner’s or an AI’s opinion can raise a priority, never lower it.
Exploit code
Akesis takes in what a penetration test found and how sure it is, never the exploit itself: that stays in your CI.
Your data and AI
Never used to train models. AI requests for EU customers run on Amazon Bedrock in the EU. Data is stored in the US; EU data storage is coming soon.
Isolation
Each organisation’s data is separated by the database itself, not only by application code.

Questions, answered plainly.

Where is my data stored?

In the United States (AWS us-east-2), in one database where each organisation is kept separate by the database itself. AI requests for EU organisations run on Amazon Bedrock in the EU. EU data storage is coming soon; tell us if you need it.

Which scanners does it work with?

Dependabot, Snyk, Strix (AI penetration testing) and any scanner that writes SARIF, such as Trivy, Grype, OSV-Scanner, CodeQL, Semgrep and Checkov. It adopts fix pull requests from Dependabot, Snyk and Renovate. Wiz and Amazon Inspector come next.

What about servers and network devices?

For any CVE on a finding, Akesis shows the vendor’s own fix: the Microsoft KB update, the Red Hat, Ubuntu or Debian package version, the PAN-OS release, the Cisco security advisory for the affected product family, and CISA’s required action. Cisco’s fixed release depends on each device’s version, so Akesis links to Cisco’s fixed-software guidance rather than guessing. It doesn’t install patches on your machines.

Does it work with Jira and Slack?

Yes. Findings that can’t become a pull request can open Jira issues, by hand or automatically, and Akesis closes them when the scanner confirms the fix. Slack gets alerts and a daily digest through a webhook you create.

Will it merge code without asking?

Only in a repository you have set to T3, and only for registry fixes that pass every gate. New repositories start at suggest-only, and an emergency stop is always one click away.

Does our source code go to an AI model?

Not to rank or fix dependency alerts. AI writes optional plain-English risk narratives from the finding’s details, never your source code. Requests run on Amazon Bedrock in your region and are not used for training.

What access does it need?

On GitHub, an app installed on the repositories you choose; on GitLab and Bitbucket, a group or workspace access token you create. It reads alerts, code and CI results, and opens and merges pull requests. Nothing else.

How do people sign in?

Single sign-on through your identity provider over OIDC: Okta, Microsoft Entra ID or Google Workspace.

Can we see everything it did?

Yes. Every action is in an audit log you can filter, export and verify independently of Akesis.

Can I ask an AI assistant about our findings?

Yes. Akesis is a read-only MCP server: connect it to Claude, sign in with your Akesis account, and ask about your backlog, a finding or a fix in plain language.

Bring the backlog.

We’re working with a small group of design partners. Bring your scanners, your repositories and your questions.

Book a call