Codebahn

Security

Last updated: 16 November 2026

EU-only. No asterisks. Codebahn is a Swedish company running on EU infrastructure with every sub-processor EU-incorporated. No data leaves the EU for storage, processing, or analytics.

Where your code lives

Primary compute and storage on Scaleway in Paris, France. Encrypted backups on Hetzner in Falkenstein, Germany. Different provider, different region.

See also: sovereignty and data residency, sub-processor list.

How it is protected

Data is encrypted in transit and at rest. In transit: TLS 1.2+ for all connections, TLS 1.3 where both endpoints support it, SSH for Git. At rest: AES-256 for repositories, the database and object storage. Repository keys are ours; the managed database and object storage use the infrastructure provider's keys. There are no customer-managed keys. Backups are encrypted client-side before upload.

Encryption at rest protects data on storage media. It does not restrict what our operators can reach: we hold the keys, and the service reads your repositories to render diffs, run search and run CI (Terms 5.4). The server's operating system is outside its scope and holds no repository data.

Who can access it

Every repository on Codebahn is private. There is no public repository hosting. One tenant cannot discover that another tenant exists: cross-tenant web and REST API requests return 404, not 403. Git over SSH, LFS and package endpoints can still reveal that a repository exists to a signed-in user without access. Explore pages, user search, and organisation search are blocked.

CI jobs run in ephemeral virtual machines, not shared containers. A VM only ever runs one organisation's jobs, so a container escape compromises only that organisation's own jobs. Runner VMs accept no inbound traffic, and the cloud metadata endpoint is blocked. VMs are reused within the same organisation for up to 6 hours, keeping Docker images and volumes between jobs. After idle timeout or 6 hours, VMs are destroyed. No state crosses tenant boundaries. See CI runners for specs and compatibility.

On the server side, production access is separated into tiers. The application runs as an unprivileged user with no sudo. We do not access repository content unless required for a specific support request, security incident, or abuse investigation. Server logs containing IP addresses are retained for 30 days, then deleted. Your own security log (sign-ins, credential and account changes) is retained for 90 days, and if you belong to an organisation, its audit log (member and permission changes, repository settings) is retained for one year. See the audit log for what each log records and how to export it.

For data privacy and what we collect, see Your code stays private.

Backups and recovery

Daily encrypted backups to a separate provider and region. Verified automatically every week: the latest snapshot is restored to a temporary directory, PostgreSQL dumps are validated, and a random sample of repositories is checked with git fsck.

Deploys use blue-green switching with health checks. A pre-deploy snapshot of git data is taken before every release, so rollback does not depend on the backup schedule.

Backups protect against data loss, not against the founder becoming unavailable. For what happens then, see Continuity.

Incident response

When something goes wrong, we follow a structured process: detect, classify, fix, communicate, review.

SeverityDefinitionResponse target
CriticalData breach, unauthorised access, complete outageWithin 4 business hours
HighPartial degradation affecting multiple customersWithin 1 business day
MediumLimited impact, single-customer issueWithin 1 business day
LowCosmetic or informational, no direct user impactWithin 5 business days

Business hours are 08:00 to 17:00 CET, Monday to Friday, excluding Swedish public holidays. Nobody is committed to watching alerts outside them, so an incident that starts then may first be picked up on the next business day.

Critical and High incidents are communicated via email to affected customers and updated on the status page. For incidents involving personal data, we notify affected data controllers within 48 hours per our DPA and GDPR Article 33, or within 24 hours on Assurance.

Every Critical and High incident gets a post-incident review within five business days, covering timeline, root cause, impact, and prevention. We publish a summary for incidents with broad user impact.

Vulnerability disclosure

If you find a security vulnerability, email security@codebahn.net with a description, steps to reproduce, the affected component, and your severity assessment.

In scope: codebahn.net (Forgejo instance, API, Git endpoints, documentation), CI runner infrastructure, and the Codebahn CLI/MCP server. Out of scope: upstream Forgejo vulnerabilities (report to the Forgejo security team), third-party services, social engineering, and physical attacks.

Our targets: these are aims, not contractual commitments.

StepTarget
AcknowledgementWithin 2 business days
Initial assessmentWithin 5 business days
Fix deployed (critical/high)Target within 7 days
Fix deployed (medium/low)Within 30 days

Safe harbour: If you act in good faith, we will not pursue legal action. Good faith means you do not access other users' data, do not degrade the service, stop and report once you have enough to demonstrate the vulnerability, and give us reasonable time to fix before public disclosure.

Certifications

We do not hold ISO 27001 or SOC 2 today. If your procurement requires signed audit reports today, we are not your vendor yet. We are working towards ISO 27001, scoped to the Codebahn service and how we run it:

We add the date to each step when it is done. Until a certificate is issued, the security addendum, NIS2 assessment and DORA register are what we can give your review.

Our infrastructure vendors are independently certified:

These certifications cover the infrastructure and payment systems that Codebahn runs on. They do not cover Codebahn's own operational processes.

We do have a Data Processing Agreement that follows GDPR Article 28, with audit rights, 48-hour notice of breaches and security incidents, and 30 days' notice before a sub-processor is added or replaced, with a right to object.

For procurement and compliance

One-click export in standard Forgejo format. The archive includes all repositories, issues, pull requests, labels, milestones, releases, and reviews. No ticket required, no waiting period. See exporting your data.

For security questionnaires or vendor assessments, email security@codebahn.net. If you are evaluating us, your first security questionnaire is answered free of charge. Recurring questionnaires, and written commitments rather than an audit report, are what Assurance adds: the addendum and register below, 2 answered security questionnaires a year, a yearly signed attestation letter, and a named contact.

14-day money-back guarantee.