Label2Prod

Security

How Label2Prod handles your code and keys

Label2Prod needs access to your issues and your code. This page says exactly what it reads, what it writes, where your code is worked on, and what stays under your control.

Read the details Ask a security question

At a glance

Built to pass a security review

Your Anthropic account

The AI works in your own Anthropic account, under your agreement with Anthropic, which bills that usage to you.

Isolated and short-lived

Every attempt gets a fresh sandbox that holds no keys or tokens and is deleted when the attempt ends.

Least access

It reads only the issues you label and pushes only to its own branches. It will not change your pipeline settings, secrets or key files.

People merge

Label2Prod never merges. Your reviewers approve every change, under your own branch protection.

Encrypted and kept apart

Credentials are encrypted with AES-256-GCM, and row-level security keeps each organization’s data apart.

A complete trail

Every update is a new Jira comment and every revision a new commit. Nothing is edited or rewritten.

Architecture

Where your code goes

Label2Prod coordinates. The AI works in a sandbox in your Anthropic account. Only people merge.

Jira Cloud

Your issues

  • Reads the issues you label, in projects you switch on
  • Writes new comments and never edits them
  • Access through Atlassian’s consent screen

Label2Prod · Frankfurt, Germany

Coordinates, never merges

  • Holds your credentials, encrypted with AES-256-GCM
  • Pushes only to its own branches
  • Keeps each organization’s data apart

GitLab

Your code

  • A Developer token that you create
  • Merge requests for your people to approve
  • Your branch protection still applies

Your Anthropic account

A sandbox for each attempt

  • Fresh for every attempt, deleted afterwards
  • Holds no keys or tokens and cannot push
  • Reaches only Label2Prod and public package registries

The details

Everything, in writing

Where your code is worked on

  • Each attempt runs in a fresh, isolated sandbox hosted by Anthropic in your own Anthropic account. The sandbox is deleted when the attempt ends.
  • The sandbox holds none of your keys or tokens and cannot push to your repository. It can reach only Label2Prod and public package registries.
  • Before any merge request opens, a second, separate sandbox checks the change on a clean copy of your repository.
  • Only Label2Prod’s own servers push to GitLab, and only to branches that start with autofix/.

What Label2Prod reads

  • Jira: the summary, description, components and comments of the issues you label, in the projects you switch on. Attachments are not opened.
  • GitLab: the repositories in the group you connect, to work on the one an issue belongs to.

What Label2Prod writes

  • Jira: new comments on the issue, a link to the merge request, and the trigger label when you start or stop a fix from the app. It never edits or deletes an earlier comment.
  • GitLab: branches that start with autofix/, commits on them, merge requests, and a verification status on those commits.
  • It never merges. There is no merge step in the product: your people review and merge.

The access it asks for

  • Jira Cloud connects through Atlassian’s consent screen with the scopes read:jira-work, write:jira-work, manage:jira-webhook and offline_access. Connect with a service account, so comments do not carry a person’s name and access does not end when someone leaves.
  • GitLab connects with a token that has the Developer role and the api scope, which lets Label2Prod push branches and open merge requests. Keep your default branch protected: with protection on, a Developer token cannot push or merge to it.
  • Anthropic uses an API key you create. Create a separate workspace for Label2Prod in the Anthropic Console and give it a spend limit.

How keys and tokens are stored

  • The credentials Label2Prod uses to reach Jira, GitLab and Anthropic are encrypted with AES-256-GCM. Only the part of Label2Prod that calls those services can decrypt them.
  • The app never shows your Anthropic key again; you see only its last four characters.
  • Each organization’s data is kept apart in the database by row-level security.
  • All traffic is encrypted in transit.

Your Anthropic account

The AI work runs under your Anthropic account and your own agreement with Anthropic, which bills that usage to you. Each attempt stops at a fixed spend ceiling on your key, and you set the monthly limit in the Anthropic Console. Label2Prod does not use your code or issues to train AI models.

Where Label2Prod runs

The Label2Prod application and its database are hosted in Frankfurt, Germany. Anthropic processes the work of each fix in its own regions, under your account. The full list of providers and what each receives is in the privacy policy.

Keeping and deleting data

Sandbox copies of your code are deleted when each attempt ends. Your connections and fix history are kept while your organization exists, so you can see past fixes. Write to hello@label2prod.com to have your organization’s data deleted. What Label2Prod already wrote to Jira or GitLab stays there, under your control.

Reporting a vulnerability

Found a security issue? Write to hello@label2prod.com. We read every report and reply as soon as we can. Our security.txt has the same contact.

Start with your next Jira issue

Label2Prod is free for early-access teams, and you don’t need to be a developer. Connect Jira and GitLab, add the label, and approve the tested result.

Start free About early access
  • Free during early access
  • No credit card
  • Nothing to install