AgentsInFlow

1 review

Self-hosted workspace for governed AI development/automation, accessible from desktop, tablet, and phone. Run Claude Code, Codex, Cursor, and OpenCode in isolated runtimes with persistent memory, ticket-driven orchestration, and full session history. Each engine operates in its own sandboxed environment with git worktree branching, browser access through an extension bridge, and voice-to-text input. Multi-project management with independent ticket hierarchies, folder organization, and cross-project orchestration. Session forking, keyboard shortcuts editor, command center, and a 3-level settings cascade for teams. Supports 28 diagram syntaxes (Mermaid, PlantUML, D2, and more) plus freehand sketching and drawing. Persistent project memory with scoped recall across assistant, project, and execution levels. Dynamic dashboards with agent-generated metrics and scheduled reporting. Full observability across agents: activity logs showing which agent executed which ticket, parent-child orchestration chains, and session transcripts. Agents can run on schedules, produce status dashboards, and surface trends without manual intervention. Free during early access, no sign-up, no credit card.