A self-hosted AI workspace built around context management — works solo, scales to teams. What makes loopat different is the context architecture: chat, code, memory, and knowledge interlock so context doesn't get lost across sessions or teammates.
When humans collaborate with AI, three things only humans can bring: drive, attention, and entropy reduction. loopat builds these in as first-class concepts — Loop, Focus, and Context — with Chat coordinating the team on the sync axis.
Every loop runs in its own sandbox with a versioned toolchain and pinned credential vault. Same state tomorrow, different machine.
WebSocket-powered real-time chat with Claude. Multiple browser tabs can share the same loop session simultaneously.
Built-in Kanban with drag-and-drop, real-time sync across subscribers. Manage tasks alongside your AI agent.
Full PTY terminal inside the sandbox. Shared across loop subscribers β everyone sees the same bash session.
Browse and edit files inside the sandbox from the web UI. Git integration with commit and log views.
Two-tier memory system — personal (personal/memory) and team-shared (context/notes/memory). Auto-recalled across sessions so context is never lost.
Every loop bundles its own context, AI agent, and working directory inside a Linux mount namespace. Portable, reproducible, and secure by default.
Knowledge + Notes + Memory
Claude CLI in bwrap
Git worktree per loop
Isolated & reproducible
Hono router binds requests to loop IDs via closures
Per-loop session maps prevent cross-loop broadcast
Each loop spawns independent Claude CLI subprocess
Kernel-enforced filesystem isolation via bubblewrap
Same workspace whether you're alone or onboarding teammates. Solo, it's a personal AI workspace; with a team, shared knowledge and notes repos sync across members.
| Claude Code | Cursor | opencode | Codex | loopat | |
|---|---|---|---|---|---|
| Form factor | CLI | IDE | TUI | Web (hosted) | Web (self-hosted) |
| License | proprietary | proprietary | MIT | proprietary | Apache 2.0 |
| Team IM integration | manual paste | manual paste | manual paste | manual paste | built-in |
| Shared team knowledge | individual config | individual config | individual config | individual config | git-synced |
| Per-session sandbox | process-level | process-level | process-level | managed | bwrap + Docker |
| Credential isolation | shared env | shared sub | shared env | acct-managed | per-loop vault |
| Data location | local files | cloud | local files | OpenAI servers | git repos you control |
| Agent engine | Anthropic | multi-model | pluggable | OpenAI | Claude Agent SDK |
No database. Everything is a file. rsync your workspace to another machine and pick up exactly where you left off.
Sandbox permissions arenΚΌt hidden in config files. TheyΚΌre derived from the physical structure of loops/<id>/ β run ls -laR to see the full access scope.
rsync $LOOPAT_HOME to any machine, restart the server, and everything works β chat history, sessions, memory, sandbox views. All paths are virtual.
All artifacts live in plain git repos you fully control; vault secrets are git-crypt encrypted. BYO API key — nothing leaves your machine except the model API call itself.
One command to start. Works on any Linux machine. macOS via Docker.