Highlights
- Use your agent through chat. Some laptops are locked down: nothing to install, connectors are admin-only, and the assistant can only fetch a web page. ProductBrain now works there too. Click Agent relay in the top bar, start a chat session, and paste the starter prompt into your assistant. It reads your plan through a private link, answers with its changes as text, and you paste them into ProductBrain. You see exactly what will change before you apply it, then paste the reply back. The link expires in 30 days, or revoke it whenever you like. How it works.
- Paste in a plan. If your agent drafted a plan somewhere else, click Import plan in the top bar and paste it in, as PLAN.md or JSON. You get a preview of what will be added before anything changes. No connector, no install.
- Point a node at your tracker. Any goal, need, bet, job or task can carry up to three links to another system, such as a Jira issue or a Linear project. ProductBrain keeps the pointer and leaves the ticket where it lives.
Other improvements
- A bet can now serve more than one need. It still sits under one need, and it can list the others it also helps.
- Needs can hold evidence of demand: who felt it, how much is behind them, and when. It is shown next to the need and never used to rank one need above another.
- Try it without an account: the demo gives your own agent a temporary plan to build while you watch it appear.
- A new project starts with a choice: use your own agent, use your agent through chat, or start a guided chat with ours.
- Removing a dependency from a bet in the side panel now sticks after a reload.