Skip to content
r:re:view
The cast Pricing Security Sign in Install on GitHub

Security

Security & data handling

What Re:View can see, where your code goes during a review, and what we keep afterwards. Written for whoever has to approve installing a code reviewer.

Last updated: November 21, 2026

On this page

  1. In short
  2. What the App can access
  3. How a review runs
  4. AI processing
  5. What we store
  6. Sign-in and sessions
  7. Billing data
  8. Infrastructure
  9. Subprocessors
  10. Removing your data
  11. Reporting a vulnerability

In short

  • Re:View reads code and posts reviews. It cannot push commits, create branches or merge pull requests.
  • It only sees the repositories you select when you install the GitHub App.
  • Each review runs in its own single-use container with a read-only filesystem, no privileges and tightly restricted network access. The container and the checkout are deleted when the review ends.
  • To write the review, parts of your code are sent to the third-party AI model providers listed below.
  • We do not keep copies of your repository. We keep review metadata (which pull request, which reviewer, when, how many tokens) for usage, billing and support.

What the App can access

Re:View is delivered as the Re:View Bot GitHub App. When you install it, GitHub asks you to choose all repositories or only selected ones. Re:View can access only the repositories you choose, and you can change that list at any time in GitHub under Settings → Applications.

The App requests these permissions:

GitHub App permissions requested by Re:View Bot
PermissionLevelWhy
ContentsReadFetch the pull request's code and your .re-view.yml config.
Pull requestsRead & writeRead the pull request and post reviews and inline comments.
IssuesRead & writePost conversation comments on pull requests (GitHub handles these through the Issues API) and read @re-view commands.
ChecksRead & writePublish the Re:View approval check on pull-request commits, which your branch rules can require.
MetadataReadBasic repository information, required by GitHub for every App.

The App receives pull request, pull request review, issue comment and installation events, so it knows when a pull request is opened or updated, when someone approves or requests changes, when someone comments @re-view, and when you change which repositories it can see.

Re:View does not have write access to repository contents, so it cannot change your code. On small, documentation-only pull requests it may post an approving review instead of running the full cast. Your branch protection rules decide whether that approval counts toward merging.

How a review runs

  1. Fetch. For each review, we fetch the pull request's head and the default branch into a fresh working directory on our review server, using a short-lived GitHub installation token. The token is handed to git for that one command and is never written into the checkout.
  2. Sandbox. The reviewer runs in a new container created for that job alone:
    • your code is mounted read-only, and the container's own filesystem is read-only apart from a small, size-capped scratch area in memory;
    • it runs as a non-root user with all Linux capabilities dropped, privilege escalation disabled, and CPU, memory and process limits;
    • it sits on a private per-job network with no direct route to the internet. It has two ways out: an egress proxy that allows HTTPS to GitHub's API and nothing else, and a credential proxy that forwards requests only to the AI model provider. The provider's API key stays in that proxy and never enters the reviewer's container.
  3. Post. Before the review is posted, we scan the reviewer's output for our own credentials and remove any we find.
  4. Destroy. When the job ends, whether it succeeded, failed or timed out, the containers and the per-job network are removed and the working directory holding your code is deleted. Every job has a hard time limit.

We treat your pull request as untrusted input. The sandbox is built so that a malicious change cannot reach the internet, read another customer's code or obtain our credentials.

AI processing

Reviews are written by large language models run by third-party providers. To produce a review, we send the provider:

  • the pull request's diff and the full contents of the changed files (within size limits), together with the reviewer's instructions;
  • other files from the repository that the reviewer reads for context during the review.

Before the full review, a smaller, faster model picks which specialists the pull request needs. That step receives only the list of changed file paths, not file contents.

Reviews are currently generated with Anthropic Claude models, and the triage step uses a Google model. Both are reached through OpenRouter. The providers process this data to produce the review and return it to us.

XL8 does not use your code to train models. Each provider's own retention and use of the data it receives is governed by its API terms, not by this promise.

What we store

We do not keep copies of your source code after a review. We do store the following in our database:

  • Installation: GitHub installation ID, account or organization login, plan, settings and install date.
  • Repositories: repository full name, default branch, whether reviews are enabled, and the contents of your .re-view.yml config file with the time it was last fetched.
  • Review jobs: pull request number, head commit SHA, which reviewer ran and what triggered it, status, an error message if the job failed, the ID of the review posted on GitHub, start and finish times, input and output token counts, and the model identifier.
  • Approval requirements: for repositories using the Re:View approval check, how many human approvals each pull-request commit needs and a short reason (for example "touches billing"), kept for 180 days.
  • Usage: monthly review counts, used to show usage and apply plan limits.
  • Billing references: plan, seat count, subscription status, and Stripe customer and subscription identifiers.
  • Delivery bookkeeping: GitHub webhook delivery IDs, so we ignore duplicates, and a record of one-time notices posted on a pull request (repository name, pull request number and notice type), so the same notice is never posted twice.

The reviews themselves live on GitHub, in your pull requests. Our servers also keep operational logs, such as job timings and errors, which we use to run and debug the service; these are kept for up to 30 days.

Records are deleted on a schedule: review job records after 13 months, one-time pull request notices after 180 days, and webhook delivery IDs after 30 days. Monthly usage totals and billing records are kept while the App is installed.

Sign-in and sessions

You sign in to the dashboard with your GitHub account (OAuth). We use the GitHub access token only during sign-in, to look up your username and the Re:View installations you can access, and we do not store it. Your session is a signed cookie holding your GitHub username and those installation IDs. It is HTTP-only, sent only over HTTPS, and expires after 24 hours. Signing out deletes it.

Billing data

Paid plans are sold through Stripe and its Link service as merchant of record. You enter card details on Stripe-hosted pages and they go directly to Stripe; Re:View never sees or stores full card numbers. We keep only Stripe identifiers and subscription status. You manage payment methods, invoices and cancellation in Stripe's customer portal.

Infrastructure

  • The review service and its database run on AWS in the US East (N. Virginia) region.
  • Data at rest is encrypted, including our database and the servers where reviews run. The database is replicated across multiple availability zones, and connections to it require TLS. Service secrets, such as the GitHub App private key, webhook secret and Stripe keys, are kept in AWS Secrets Manager.
  • Every GitHub and Stripe webhook is checked against its signing secret before we act on it.
  • DNS is provided by Cloudflare.

Subprocessors

These providers process customer data on our behalf. The authoritative list is on our Subprocessors page, and we give at least 30 days' notice before adding or replacing one: on that page, by email to the billing contact of every paid installation, and by email to anyone who asks to be on our notice list. Our Data Processing Addendum covers how they are bound.

Subprocessors used by Re:View
ProviderPurposeData involved
Amazon Web ServicesHosting, database and secrets (US East)The service data described on this page
OpenRouterRouting requests to AI modelsCode, diffs and file paths sent for review
AnthropicAI model that writes reviewsCode and diffs sent for review
GoogleAI model that selects reviewers (triage); email hosting for contact@re-view.aiChanged file paths; emails you send us
GitHubApp platform, sign-in, and where reviews are postedRepository access, account identity
Stripe (including Link)Payment processing and merchant of record (reseller) for paid plans, including taxBilling contact, tax and payment details
CloudflareDNS, the www redirect, and forwarding of email sent to contact@re-view.aiIP address and website requests; emails you send us

Removing your data

When you remove a repository from the App's access, we delete that repository's records and its review history. When you uninstall the App, we set any paid subscription to cancel at the end of the current billing period, then delete the installation and everything tied to it: repositories, review jobs, usage records, and our subscription records. Reviews already posted on GitHub stay in your pull requests until you delete them there. Stripe keeps its own billing records under its policies.

To ask us to delete anything else, email contact@re-view.ai.

Reporting a vulnerability

If you think you have found a security issue in Re:View, email contact@re-view.ai with "Security" in the subject, a description of the issue and steps to reproduce it. We will acknowledge your report within 3 business days and keep you informed as we investigate and fix it.

Safe harbor

We will not pursue legal action against, or ask law enforcement to investigate, anyone who researches and reports a vulnerability in good faith under this policy. Good faith means that you:

  • test only against your own installations and accounts, and do not access, modify or keep other customers' data; if you reach someone else's data by accident, stop and tell us;
  • avoid degrading the Service for others, including denial-of-service testing, spam and social engineering of our staff or customers;
  • give us reasonable time to fix the issue before disclosing it publicly, and coordinate the disclosure date with us;
  • do not ask for payment in exchange for the details of a report.

Activity consistent with this policy is authorized for the purposes of applicable anti-hacking laws, and we will not treat it as a breach of our acceptable use terms. We do not currently run a paid bug bounty. Our contact details are also published at /.well-known/security.txt.

r:re:view © 2026 Re:View · re-view.ai PrivacyTermsSecurityDPASubprocessorscontact@re-view.ai