Barbequeue

• 

  Product Landing Page

Inspiration

AI planning is powerful, but most of it is still single-player. One person works with the agent, while the actual team alignment happens later in Slack, meetings, or scattered comments. Barbequeue was inspired by the idea that good plans should be grilled by the whole team, not just polished in private.

What it does

Barbequeue is a multiplayer AI grilling session for software teams. A host starts a session from their Codex environment, keeps repository access under control, and invites teammates into a browser-based participant room. The AI asks grilling questions, participants contribute ideas, clarifications, objections, and abstentions, and the group converges on accepted outcomes. At the end, the host gets a session record and a documentation change ready to review.

How we built it

We built Barbequeue by extending the existing grill-with-docs workflow into a hosted multiplayer experience. The host keeps raw repository files local, while participants only interact with shared, approved session context. The alpha includes the core session flow: host-led grilling, participant rooms, contributions, objections, answer candidates, consensus states, and documentation output. For the hackathon presentation, we also built a React + Vite pitch landing that works as a stage deck and visual demo, showing the core mechanism: many human inputs resolving into one shared decision.

Challenges we ran into

The biggest challenge was making multiplayer AI feel structured instead of turning it into another chat room. We had to model contributions, objections, abstentions, answer candidates, and accepted outcomes as distinct parts of a decision workflow. Another challenge was trust. Participants need enough context to contribute, but the host must stay in control of repository exposure. Designing that boundary clearly was central to the product.

Accomplishments that we're proud of

We are proud that Barbequeue is already working in alpha, not just as a concept. It takes a real single-player agent workflow and turns it into a collaborative session where the team can challenge, refine, and accept decisions together. We are also proud of the product shape: objections are not treated as noise, consensus is explicit, and the output is useful documentation rather than just a meeting summary.

What we learned

We learned that multiplayer AI is not about adding more people to a chat with an agent. It is about creating a shared decision structure where people can safely disagree, resolve objections, and leave with a trustworthy record. We also learned that repository context changes the collaboration model. Privacy, host control, and selective context sharing are not edge cases; they are core requirements.

What's next for Barbequeue

Next, we want to harden the alpha through real team usage. That means improving session reliability, refining participant interactions, tightening the consensus flow, and making the documentation output more useful for different repositories and planning styles. We also want to explore deeper Codex integration, better async participation, and richer session records so Barbequeue can become the default way teams grill important technical decisions together.

Built With

• cloudflare
• codex
• react
• websockets