Security
Security overview
FT2's Host-first architecture, security boundaries, and vulnerability-reporting process.
1. Security model
FT2 keeps the authoritative source for projects, threads, agent configuration, CLI credentials, and execution state in the user-controlled Environment. The public Gateway provides authenticated registration, routing, and streaming; it is not designed as a model proxy or project database.
2. Connections and pairing
- Public connections use HTTPS/WSS;
- The Host opens an outbound connection to Gateway, so a Host management port does not need to be public;
- Pair Tokens are short-lived, one-time tokens bound to device registration;
- Environment identity and authorized devices are kept in the Host's persistent state directory.
3. Human in the loop
Commands, file changes, and requests for additional permission can be explicitly approved or rejected on mobile. This reduces invisible execution but does not replace least privilege on the Host, code review, testing, backups, or production change controls.
4. Current boundaries
Terminal and agent processes ultimately run with the Host user's permissions. FT2 does not automatically reduce operating-system privileges. Use a dedicated account, grant the smallest practical root directories, and restrict high-risk credentials.
5. Release security
Public desktop and Server downloads will open after code signing, version manifests, SHA-256 verification, atomic upgrades, and rollback procedures are stable. Internal test artifacts should not be treated as production releases.
6. Report a security issue
Email suspected vulnerabilities to [email protected] or [email protected] with “FT2 Security Report” in the subject. Include reproduction steps, scope, and redacted evidence. Do not attach real tokens, API keys, user project code, or third-party credentials.