Security

Please report vulnerabilities privately to hello@jonna.dev. Don't open a public issue for a security problem.

What to include in a report

• The version (Help, then About)
• Your operating system
• The steps to reproduce
• What an attacker could do

We aim to reply within a week, and to tell you when a fix ships. Please give us a reasonable time to fix it before telling anyone else. There's no bug bounty yet.

How the editor is designed

The desktop app uses Tauri and Rust. Search, file access, terminal processes and other native work run behind typed commands. The agent and extensions use the editor's command system, where permission checks and audit records apply.

Review and recovery. Agent edits can be reviewed per file or per hunk before they are applied. Per-turn checkpoints, local history and workspace undo help you recover earlier work. Command execution has configurable approval modes; review those settings before granting broad access.

Extension permissions. Third-party extension code runs in workers with a permission-checked API. File access, shell commands, network requests and other capabilities require the permissions you grant. An extension created by an agent follows those permission checks too.

Models and integrations. Local models and hosted models have different data boundaries. Hosted providers receive the context sent to them. MCP servers and external agents have their own tools and permissions, so choose and configure the integrations you trust.

Accounts

Signing in from the app uses OAuth 2 with PKCE through your own browser, so the app never sees your password. Sign-in codes are single use and expire in two minutes, and our servers store only hashed copies of session tokens.

Supported versions

The latest release. Fixes ship as a new release.