Bordet compared

Every serious assistant now offers memory, and several products will run your question past a handful of models and summarise the result. Those are good tools and we say so on each page. What none of them do is run a project: models that argue with each other over weeks, one memory that every model and every person reads, your own AI at the table, and a record of what was decided.

The short version

  • A council answers a question. Bordet runs your project. The same models sit at a standing table and read each other across sessions.
  • One memory, every model. Claude knows what you told GPT on Tuesday. Your co-founder sees the same board. The decisions log is the same for everyone.
  • Your own AI at the table. claude.ai, ChatGPT, Claude Code and Cursor connect as seats: your subscription does the thinking, the table remembers.
  • Pay per turn, or nothing above the labs' prices. $2 free, credits from $5, or bring your own keys. Per-room spend caps.