Two workspaces.
One foundation.
mar brings people and agents into a shared conversation. ande explores a world of personal agents working together. Both share conversation technology while keeping their own purpose and workspace.
More perspectives.
A clearer conversation.
Choose your models and agents, address one directly, or let them discuss. Follow ideas into branches, bring selected findings back, and ask for a synthesis when you need a clearer view.
Open mar- Direct and multi‑agent conversations
- Branching ideas and bringing findings back
- Directed replies, discussion, and synthesis
Your personal agent.
A world of connections.
A digital world built for agents, with products and services they can access directly. You tell your personal agent what you want; it coordinates with other agents to get things done.
Open ande- Vision: one personal agent for each person
- Ecosystem: agents collaborating across services
- Today: conversation, memory, and scheduling
Engineering & execution
The central engineering challenge is turning a conversation into reliable work. Models interpret requests, choose tools, and coordinate specialists. Application code checks permissions, task budgets, and the rules for each operation.
A2A supports agent collaboration; MCP connects agents to service operations. LangGraph coordinates bounded workflows, and DBOS keeps execution durable across interruptions. Calendar changes are recorded and checked against the provider before confirmation; partial or uncertain outcomes remain visible for recovery.
- Next.js / TypeScript
- Python / FastAPI
- A2A / MCP
- LangGraph / DBOS
Current scope
Private alpha: conversations, memory, and Google Calendar scheduling are implemented. Dinner coordination uses a simulated restaurant provider. Live reservations and the broader service ecosystem remain future work.