03
Athena assistant
AI that takes action, after checking the risk tier: low-risk work runs, irreversible work waits for you.
Athena is the assistant inside the platform. It converses, remembers what you allow it to, runs on schedules, and completes work through skills and tools. Every action is risk-tiered to decide whether confirmation is needed.
- 03.1
Conversation
Talk to the assistant on Web, Desktop, LINE or a public link.
- 03.2
Profiles & skills
Admins configure agent profiles for each purpose and attach skills and skill packs.
- 03.3
Memory & schedules
Users review and manage their own memories; scheduled jobs leave run records.
- 03.4
Tools & devices
MCP servers and controlled device capabilities extend what it can do; high-risk capabilities are off by default.
Risk tiers: confirm what matters, not everything
Every assistant action is classified before it runs. The tier decides whether a person must confirm.
- R0
Read-only
Queries and explanations run directly.
- R1
Reversible
Writes that can be undone in the product run directly and report back, with undo where needed.
- R3
Irreversible
Hard deletes, permission and key changes require strong confirmation or step-up, and approval by another role where needed.
R2 has been merged into R1; tier numbers are kept so records stay consistent.
- Agent workbench
- Build, test and publish your own agents.
- Agent registry & assignment
- Admins manage agents and who they are assigned to.
- Skills & skill packs
- Import, request and review skills.
- Capability proposals
- New capabilities go through review before they take effect.
- MCP servers
- Connect external tools within granted scope.
High-risk capabilities are off by default. A capability existing in the product does not mean it is enabled in your environment.