# Trust — Cerebe. Cerebe is the quality engine for AI generated code

Where repository content goes, how review agents are isolated, and how to verify a signed verdict. Cerebe is the quality engine for AI generated code

[Cerebe](/) / Trust

# Who reviews your code and what they can access.

[Install the GitHub App](https://github.com/apps/cerebe-factory/installations/new) [Request security documents](#documents)

Setup instructions

Paste this prompt into your coding agent. [Full installation guide →](/install)

```
Install Cerebe in this repository. Follow https://cerebe.ai/setup.md in order.
Lead with the GitHub App at https://github.com/apps/cerebe-factory/installations/new, then the CLI, and finish by running cerebe login.
When cerebe is already on PATH, run cerebe --help. If login is not listed, run the installer before cerebe login.
When login is listed and this checkout has no cerebe/ directory, run cerebe install instead of the installer.
When this checkout already has a cerebe/ directory, run cerebe init.
```

Independence

### The reviewer is not the author.

Agents from competing model vendors review every change. The check is posted by Cerebe, outside the platform that wrote the code.

Isolation

### Sandboxed, with no internet access.

Each agent sees only the change and its repository. Provider credentials are never exposed to review agents.

Data

### What happens to your code.

Your change, and the repository content needed to review it, go only to the AI model providers in your review configuration, in the regions your account is configured for. The sandbox is destroyed after each review. Retention and deletion controls are not available yet.

## Control status

Control status as of
2026-10-27
.

| Control | Status | Dated |
| --- | --- | --- |

| Critics from competing model vendors | Shipped | 2026-10-27 |

| sandboxed with no web access | Shipped | 2026-10-27 |

| model keys never reach them | Shipped | 2026-10-27 |

| Verdicts are signed and independently verifiable. | Shipped | 2026-10-27 |

| one verdict per commit | Shipped | 2026-10-27 |

| The reviewer is not the author. | Shipped | 2026-10-27 |

| Retention and deletion controls are not available yet. | Roadmap | 2026-10-27 |

| Customer-owned rules and doctrine are on the roadmap. | Roadmap | 2026-10-27 |

[Control status (JSON)](/trust/facts.json)

Verifiable verdicts

## Signed and verifiable.

Verdicts are signed and independently verifiable. Each finding says why it matters and what the fix is. Findings include a file and line reference when applicable. Anyone with the CLI and read access to the repository can verify a verdict's signature and its binding to the commit. The same verdict math runs on your workstation and in the check; the hosted check is the approval of record.

Compliance

## Evidence your auditor can sample.

Every merge carries a per-commit record of who reviewed the change, under which rules, on which commit, with the findings and their evidence. That record is available today for change-management sampling. A one-command export pack is in progress.

Enterprise

## Run it where your code lives.

Deployment

### Your own cloud or private network.

Run Cerebe in your own cloud, with forward-deployed engineering. Private network and private cloud deployment are part of an enterprise engagement.

Keys

### Bring your own model keys.

Your own model vendor keys are set up as part of an enterprise engagement.

Sign-in

### Google or GitHub today. Your identity provider on request.

Sign in with Google or GitHub today. Single sign-on with your own identity provider, over SAML or OIDC, is set up as part of an enterprise engagement.

## Request security documents.

Privacy, terms, the data processing agreement, the subprocessors list and the security page are public. Tell us which documents you need for your security review. You can also email [hi@cerebe.ai](mailto:hi@cerebe.ai?subject=Cerebe%20security%20documents).

[Install the GitHub App](https://github.com/apps/cerebe-factory/installations/new) [Security →](/security)