AI is moving beyond assistants that simply answer questions. With tools like MCP, agents can now interact directly with project data and perform actions.

That makes me wonder: what should an agent-native project workspace actually look like?

For me, it would need to support:

• Agents receiving and owning project tasks
• Shared context between humans and agents
• Agents updating tasks and project records
• Multiple agents working in parallel
• Clear activity and accountability
• Human review before important work is completed

Zoho Projects is already moving in this direction with its MCP integration, which allows AI agents to access and manage project data through an LLM.

I've also been looking at Sharkly.ai, which takes a different approach by making agents participants in the project workspace alongside human teammates.

For teams already experimenting with AI in project management, what would make a workspace truly agent-native rather than simply adding AI features to an existing project-management tool?