🎌 Sencho — Personal
Your personal AI agent. Sencho manages daily workflows, maintains long-term memory, handles calendar and notes, and executes private tasks on your behalf.
- Daily workflow automation
- Persistent memory across sessions
- Calendar & scheduling
- Notes & task management
- Private — only you can access
⚒️ Takumi — Worker
The operational backbone. Takumi handles automation, builds, deploys, CI/CD pipelines, and infrastructure operations.
- CI/CD pipeline management
- Infrastructure provisioning
- Build & deployment automation
- GitOps-driven workflows
- Scoped to team/project boundaries
🏯 Bannin — Public
Your outward-facing agent. Bannin handles customer support (L1), security monitoring, and external communications.
- Customer support triage
- Security event monitoring
- External communications
- Escalation to human operators
- Public-safe — strict output policies
How They Work Together
Agents communicate via inter-agent messaging — a secure, audited message bus that allows coordination without sharing credentials or breaking isolation boundaries.
Fleet coordination means Sencho can ask Takumi to deploy something, Bannin can escalate to Sencho for decisions, and Takumi can notify Bannin when a deployment affects customers — all automatically, all logged.
Trust Model & Permissions
Each agent type operates within strict permission boundaries:
Sencho
User-scoped. Access limited to the owner's data, calendars, and personal resources. Cannot access other users' agents.
Takumi
Project-scoped. Access limited to assigned repositories, pipelines, and infrastructure. Governed by GitOps policies.
Bannin
Public-scoped. Strict output filtering. No access to internal systems beyond what's explicitly exposed. All responses reviewed by policy engine.