The answers your security team will ask for.
A security review asks the same questions of every tool. This page answers them in plain words. Each answer links to the public page that backs it. Send the link to your security team.
Data
What CodeHerder holds, and what stays on your devices.
Where does our source code live?
All plansOn the devices you register. Each task gets its own checkout there. CodeHerder keeps the record of the work: task text, comments and commit metadata. Agent output and attachments can contain file text or diffs, and CodeHerder stores those.
Source: Security: what we store →What does CodeHerder store?
All plansTask text, comments, messages, wiki pages, attachments, activity events and cost events. It also keeps the last 8 KiB of each session's terminal output, with secrets redacted, for 30 days by default.
Source: Security: what we store →How do you store our secrets?
All plansWorkspace secrets and device secrets use KMS envelope encryption, and a key-management failure blocks the write instead of storing plaintext. Integration tokens and webhook secrets use AES-256-GCM under a key from the server's environment instead of KMS. A production server refuses to start without those keys.
Source: Docs: Secrets →Can we set how long data is kept?
EnterpriseOn a self-hosted server, yes. The server deletes each class of data when its period ends. You change most periods with a server setting.
Source: Docs: Retention and data subject requests →
Access
Who can sign in, and what they can do.
Do you support single sign-on?
EnterpriseFederate workspace sign-in with your own identity provider, scoped to your verified domains.
Source: Docs: Single sign-on →Can our identity provider add and remove users?
EnterpriseAdd and remove workspace members straight from your identity provider. No manual invite, no manual offboarding step.
Source: Docs: User provisioning →How do you control what each person can do?
All plansCodeHerder checks a caller's role on every workspace request: owner, admin, or member. A member outside a workspace gets a 403, or a 404 where confirming the resource exists would itself leak information. A few public routes, such as health checks and sign-in, skip the role check by design.
Source: Security: roles →How do API keys work?
All plansEach member signs in with a ch_ key: 32 random bytes, shown once, stored as a hash. A member key carries the member's full access. Keys made for device registration are narrower, and so are SCIM and self-host download keys. Revoke a key, and every cached copy of that decision expires within seconds.
Source: Security: access control →Do session credentials expire?
All plansA task session or dev session runs on a credential with its own idle limit. Use it, and it keeps sliding forward. Stop using it, and it lapses on its own within a day.
Source: Security: access control →
Agents
Where agents run, and what they can reach.
Where do agents run?
All plansOn a device you register. The device is your own machine or a host in your AWS account. The agent works in an isolated copy of your repository there.
Source: Where agents run →How does a device connect to CodeHerder?
All plansDevices dial out to CodeHerder over a WebSocket secured with TLS. Nobody connects to your machine from the outside on CodeHerder's behalf.
Source: Security: encrypted outbound connections →How are agent runs isolated?
All plansEach task gets its own git worktree. In a container, each run also gets a private network and its own proxy as the only exit. A plain host process has no such limit. It can reach what its OS user can reach.
Source: Docs: Isolating agent runs →What can an agent reach on the network?
All plansIn a container run, the stage proxy refuses private, loopback and cloud-metadata addresses by default. A writable stage reaches only approved hosts, unless you change its mode.
Source: Docs: Isolating agent runs →What stops prompt injection?
All plansNothing stops a model from obeying text it reads. Labels and fences keep structure and show the source. Limits hold instead. Scoped git credentials and isolation cap the damage. So do strict egress and approval gates.
Source: Blog: What stops an agent from obeying text it reads? →Who decides when an agent needs a new host?
All plansIn a container run, a blocked host can become a request. A person approves one exact host, port and transport, or denies it. A request that nobody decides in 24 hours expires.
Source: Docs: Network access approvals →What secrets can an agent see?
All plansOnly an owner or admin can manage secrets. An agent can never view or manage them. You set the scope of each secret.
Source: Docs: Secrets →
AI providers
Who pays for the models, and who holds the keys.
Who pays for AI usage?
All plansYou do. You bring your own provider API key, and your provider bills you directly. CodeHerder charges only for your plan.
Source: Pricing →Does CodeHerder hold our provider login?
All plansNo. The agent on your device calls the provider you chose. CodeHerder never stores your coding agent's provider login.
Source: Docs: AI credentials on a device →
Audit and monitoring
What you can see, and where it can go.
Is there an audit trail?
All plansMost state-changing calls write a row: task moves, agent actions, member changes, cost events. The API has no route to edit or delete a row. Retention sweeps and workspace deletion remove rows. Personal-data erasure scrubs them. How long rows are kept depends on your plan's retention window.
Source: Security: audit trail →Can we send audit data to our SIEM?
StarterRender any outbound webhook subscription's deliveries as an OCSF API Activity record, ready for direct ingestion by your SIEM.
Source: Docs: Webhooks →
Change control
What stops a bad change from moving forward.
Can we require checks before work moves on?
All plansYes. A task cannot move past a required check or an unticked checklist item. The server refuses a required pass from the author of the work.
Source: Verification →Can we limit how large a change is?
All plansYes, for each stage. A stage can set limits on files, lines and top-level directories. When a change goes over, the task stays in its stage.
Source: Docs: Change shape →
Self-hosting and residency
Run it in your own account.
Can we run CodeHerder ourselves?
EnterpriseYes, on Enterprise. You run the server, the database and the devices in your own AWS account. Some traffic can still leave, such as the release check.
Source: Self-hosting →Where does our data live?
EnterpriseOn a self-host, where you run the server, the database and the devices. You choose the region of each AWS store. One setting places every AWS store in one region. You place the database yourself.
Source: Docs: Data residency →
Vulnerability reporting
How to report a flaw.
How do we report a security flaw?
All plansEmail the security contact in /.well-known/security.txt. You need no account. We acknowledge a report within 3 business days and give an initial assessment within 7 business days.
Source: Docs: Reporting a vulnerability →
Ask us about these
This page does not answer these topics. Contact us and ask.
- SOC 2Ask us
- ISO 27001Ask us
- Penetration testingAsk us
- Data processing agreementAsk us
- SubprocessorsAsk us
Send this page to your security team. For the full list of what CodeHerder stores and relays, read Security. To see what an agent runs in, read agent environments.
Get your security team to yes faster.
Send them this page. Then start a workspace and look at the controls yourself.