The product is a self-hosted agentic studio that runs parallel AI coding sessions with project memory automatically injected at spawn, solving repeated context loss and unsupervised agent actions. It provides a kanban board that spawns pre-contexted sessions, cost tracking per session, audit logs, and human approval for irreversible actions. It sells to developers and engineering teams building and operating software. It is positioned against agent wrappers by injecting local RAG memory and requiring human validation, delivered as open source Apache-2.0 with one-command install and an optional managed cloud.
Key features
- Parallel AI sessions on Claude Agent SDK
- Auto-injected project memory at spawn
- Local RAG with ONNX embeddings
- Kanban board to spawn sessions
- Permission requests as buttons
- Cost tracking per session
- Audit log per action
- Event-driven ops supervision
- Telegram HITL notifications
- Prometheus Grafana Loki observability
- Alert to diagnostic session
- Encrypted secrets as env variables
- Replayable runbooks from memory notes
- Versioned deploy and rollback
- On-demand dashboard generation
Social posts
- No social media activity detected
GTM channels
- Marketplace
- Changelog
ICP
- Software developers
- Engineering teams
- DevOps sre teams