Skip to main content

​
Highlights

  • Read-only agent seats. Give an agent a seat that can read the whole plan but cannot change or delete anything in it: a report bot, a reviewer, or a stakeholder’s assistant that answers questions about the plan. Tick Read only when you add an agent seat in Team members, or pass "role": "viewer" through the API or MCP. Every write it tries is refused with a clear reason and changes nothing. You can switch a seat between read-only and editor at any time, and its key stays the same. How it works.
  • Agents can use the words they already know. An agent used to Jira or Linear can send epic, ticket, story or sprint and get the right result, with a note saying how ProductBrain reads it: an epic is a bet, a ticket is a job, a sprint is a phase. Before, a filter like type=ticket quietly returned nothing. Words that mean different things in different tools, like feature or initiative, are left for the agent to translate. Full table.

​
Other improvements

  • The no-login demo offers a few one-click starter ideas when you want to watch your agent build a plan but don’t have an idea of your own to hand.