Code and data handling
Panolayer works with the repository you open on your Mac. When you use a model-backed feature, relevant prompts, code context, or repository metadata are sent for processing to your configured provider or Panolayer’s managed model service.
Panolayer does not use Customer Content to train its foundation models. We do not build a marketing or training database from your code.
Third-party coding agents are independent services. The prompts and code context you send to an agent are handled under its publisher’s terms and the policies of its model providers.
Agent access and file controls
Marketplace agents disclose their requested permissions before installation. Review project access, network access, and any supported login or configuration access before accepting.
Workspace and commands
Container agents with project write access work in a separate Git worktree with their own home directory. They can read and edit the files made available in that workspace and run commands, such as builds and tests, inside the container.
Your everyday checkout is separate from the agent’s workspace. Applying the agent’s changes to your branch is a separate review and approval step.
Files outside the project and existing logins
A container agent does not receive access to your entire home folder. Access to supported existing CLI login or configuration folders is disclosed separately. Optional requests can be declined.
Granting login access lets the agent read that login state. The original folder on your Mac is read-only to the container; the agent can maintain its own login state for later sessions.
Network access and connected tools
Network access is a separate permission. An agent granted full network access can make outbound internet and local-network requests, including requests to its model provider.
Connected tools may provide access to services beyond the project workspace. Review the permissions of those tools as well as the agent’s own permissions.
Locking and hiding files
In Architecture, open a file to set its access controls before starting a new container session.
- Lock File prevents AI agents from modifying the file. A locked file remains readable; use Hide from Agents when you also want to restrict reading.
- Hide from Agents omits the file from the working files and Git history available to new container agents. The file remains available to you in your local project. Existing agent sessions are unchanged.
Unsandboxed agents and host terminals
The container boundaries above do not apply to unsandboxed agents or host terminals. They run on your Mac with the access allowed by your user account and any separate restrictions. Check the execution mode before starting a run.
Verification and change review
Panolayer checks agent changes and separates new regressions from pre-existing failures. New regressions hold an agent’s managed merge request for repair. Applying the agent’s work to your branch requires your approval.
In Changes, you can inspect each agent’s worktree and review changed files from highest to lowest risk before approving or committing.
Like all security tools, it's impossible to catch every bug. A human override can apply changes when failures or unavailable checks remain; the agent cannot make that decision for you. Keep your project’s tests, application testing, and security review in the workflow.
Account and connection security
Website and API traffic use TLS. Desktop session tokens are stored in the platform keychain when available.
Stripe handles payment collection. Panolayer does not receive or store your full card number.
Support reports can include diagnostics, logs, screenshots, and attached files. Review what you include before sending a report.
Data retention and account deletion
Personal information is retained for the purposes described in the privacy policy, including providing the service, support, security, and legal requirements. Retention periods vary by purpose.
You can request account deletion in your account settings. On termination, Panolayer deactivates or deletes your account and information from active databases. Some information may be retained to prevent fraud, troubleshoot problems, assist investigations, enforce legal terms, or comply with legal requirements.
Security questions and reporting
Contact Panolayer with security questions or to report a suspected vulnerability. Include the affected workflow and steps to reproduce the issue. Do not include passwords, API keys, or private source code in an initial report.
For questions about personal information or to exercise your privacy rights, email [email protected].