Security & trust
Your source code is your most sensitive asset. We treat it that way.
MergeHarbor is designed so you can run it in your own environment, control who sees what, and prove what happened.
Architecture and deployment
- Deployment options
- MergeHarbor cloud, or self-hosted on Kubernetes (Helm) or AWS (Terraform). Self-hosted installations run entirely in your environment.
- Isolation
- Every request is scoped to one organization in the same query that loads it. Users and API keys from other organizations receive “not found”, so organization existence is not disclosed.
- Data plane
- Repository indexes live in independent cells. Workers keep repository mirrors on local disk and evict them under a disk quota.
- Review agent
- The review agent can only read the repository and commit it was asked to review, through short-lived signed tokens.
Data handling
- Secret redaction
- API keys, tokens, private keys, JWTs, passwords in connection strings, and high-entropy secret assignments are removed before content reaches a model, logs, or archives.
- No model training
- Customer code is never used to train models. Self-hosted customers can use their own model provider accounts.
- Encryption
- TLS 1.2+ in transit, with optional mutual TLS between internal services. Code-host tokens and identity-provider secrets are encrypted at rest with AES-256-GCM and rotatable keys.
- Retention
- Organizations set a retention period for reviews, findings, and usage records. Deleted organizations are purged after a grace period.
- Data residency
- Organizations can restrict inference to a region and to approved model providers.
- Data portability
- Owners and admins can export the whole organization as JSON Lines at any time, through a signed link that expires after 15 minutes.
- Payments
- Card details are entered only on our payment provider's hosted pages. MergeHarbor never receives or stores card data.
Identity and access
- Single sign-on
- OpenID Connect with PKCE. Organizations can enforce SSO, restrict sign-in to their domains, and map identity-provider groups to roles.
- Roles
- Built-in owner, admin, member, and viewer roles, and custom roles. Users can only grant permissions they hold, and an organization always keeps an owner.
- Sessions and API keys
- Sessions and API keys are stored only as keyed hashes. API keys are scoped to one organization and role, with optional expiry.
- Network policy
- Per-organization IP allowlists and session lifetimes. Authentication endpoints are rate-limited.
- Support access
- MergeHarbor staff have no standing access to your organization in our cloud. Support requests a time-limited session, read-only by default, that an organization owner must approve and can revoke; every action is recorded in your audit log under the staff member's name.
Audit and monitoring
- Audit log
- Every change, sign-in, permission denial, and export is recorded with actor, IP address, and before-and-after values, with secrets masked.
- Tamper evidence
- Audit records are append-only and hash-chained per organization, so modified, deleted, or reordered records are detectable.
- Integrity check
- Owners can verify the hash chain from the console at any time; the result shows how many events were checked and the first one that does not match, if any.
- SIEM streaming
- Push audit events continuously to an HTTPS endpoint (HMAC-signed), Splunk HTTP Event Collector, Datadog Logs, or an Amazon S3 bucket, with retries and a delivery cursor. Export as CSV or JSONL at any time.
- Operations
- Prometheus metrics, OpenTelemetry tracing, and readiness checks on every service for self-hosted installations.
Code-host integrations
- Least privilege
- MergeHarbor requests only the permissions it needs to read code and post reviews on each code host.
- Verified webhooks
- Every webhook is authenticated with the host's signature or token, and replayed deliveries are ignored.
- Self-managed hosts
- Calls to customer-supplied hosts are restricted to HTTPS, re-checked after DNS resolution, and blocked from internal and cloud-metadata addresses unless you allow private networks.
- Review integrity
- Review configuration is read from the pull request's merge base, so a pull request cannot change how it is reviewed. Text that attempts to steer the reviewer is flagged.
Compliance
MergeHarbor is built around controls aligned with SOC 2 and GDPR requirements: access control, audit logging, encryption, retention, and vendor data-processing terms. MergeHarbor is not yet SOC 2 certified; an audit is planned. We provide a data processing addendum, a list of subprocessors, and completed security questionnaires on request.
Responsible disclosure
If you believe you have found a security vulnerability in MergeHarbor, emailsecurity@mergeharbor.com with steps to reproduce and the affected versions.
We acknowledge reports within two business days and share a remediation plan within ten business days. We credit researchers who wish to be named. Please do not access other customers' data, degrade service availability, or publicly disclose an issue before it is resolved.
Need our security documentation?
Contact us for architecture details, our data processing addendum, and questionnaire responses.