Agent Brain Trust launches tool that lets AI systems consult panels of named expert personas to critique decisions and debate strategy

A new product called Agent Brain Trust (launched on Hacker News) gives AI agents the ability to summon virtual panels of real, named experts—such as a security specialist or product strategist—to review and critique their work. The system comes with 10 pre-built expert panels and can be extended with custom experts via an MCP server (a protocol that connects AI systems to external tools).

Unlike AI systems that reason in isolation, Agent Brain Trust stages structured debates where expert personas take turns critiquing the same decision—an architecture choice, a piece of writing, a product strategy—without skipping useful feedback. Each expert is mapped to a real persona card so the panel draws on genuine expertise rather than inventing credentials.

For product teams and engineers building AI systems, this shifts the workflow: instead of running ideas through a single AI model (which often rubber-stamps its own suggestions), you now get adversarial feedback that catches blind spots. A designer proposing a user flow gets pushback from a privacy expert and a performance expert simultaneously, surfacing tradeoffs before implementation.