| Key features | - One source of truth for every agent prompt, session, tool call, and skill
- Automatic harness detection and session import
- Sessions across harnesses including Claude Code, Codex, OpenCode, DeepSeek, Pi, Gemini CLI, Cursor, GitHub Copilot CLI, Droid
- Share sessions with team or via public link
- Move sessions from one harness to another
- Continue sessions on cloud
- S3-backed versioned filesystem with POSIX semantics
- Instant branches and sub-second forks for parallel agent execution
- Full environment control for agents including dependencies, secrets, runtime, and version history
| - Support for 11 data sources including PostgreSQL, MySQL, MariaDB, MongoDB, Redis, SQLite, Firebird, MSSQL, Docker Volume, Files, and Valkey
- Agents deployed close to databases for secure backups without exposing databases publicly
- Multi-backend storage support including S3-compatible providers, AWS S3, MinIO, and Google Drive
- Scheduled backups with fixed intervals or Cron syntax
- Retention policies with GFS tiers, keep-last-N, and keep-for-X-days rules per database
- Role-based access control and organization with RBAC
- OIDC and OAuth2 authentication with external identity providers
- Notifications and alerts for backup incidents with configurable severity levels
- REST API and MCP server for integration with existing workflows and AI agents
|
|---|
| What makes it different | - Wire-compatible with Git so agents already know the model
- Treats every action as part of a swarm, not like a human waiting at keyboard
- Reads harness stores after the fact to avoid impacting agent performance
- Portable across harnesses, not trapped in one tool
- Version control first, agent-native by design, Git-compatible at surface
| - Quick installation and deployment in under 2 minutes with a single command
- Full data sovereignty with self-hosted deployment on your own infrastructure
- Secure-by-design architecture where the dashboard never contacts agents, requiring no inbound ports
- Unified backup and restore experience across relational and document-oriented databases
|