Vercel adds Ampersand group conversation model to AI Gateway

A model designed for group conversations

Ampersand is built for settings such as team chats, where several people may have different goals and a useful assistant needs to judge when to intervene. According to Vercel, the model is designed to understand individual and shared goals, help people work through disagreements, and recognize when a contribution would be useful.

That contribution could involve asking a clarifying question or sharing relevant context. Ampersand can also wait while people respond to one another, rather than adding a message to every turn. This makes its behavior different from a conversational integration that posts a response after every incoming message.

Developers control the model’s actions

Applications expose Ampersand’s two possible actions as caller-defined function tools. The send_message tool delivers a contribution to the group conversation, while pass keeps the conversation quiet until the next human message. The application executes both tools and records their results.

Vercel’s example uses the AI SDK and the model name humansand/ampersand-v0.1. Developers can also call Ampersand through the OpenAI-compatible Chat Completions API or a coding agent. Vercel says tool choice should remain at its default, because Ampersand accepts auto and none, but rejects required and named tool choices.

The example processes one completed action per human message. It records the assistant response before delivering a validated message, and uses delivery receipts to help prevent duplicate delivery. For deployed chat applications, Vercel recommends durable storage for the response, history, and delivery receipts, sequential processing per conversation, retries for uncertain delivery using the saved action and same receipt, and an access check before posting.

Availability through AI Gateway and coding agents

Ampersand is available on AI Gateway for free through November 13, 2026. Developers can try it in the model playground. To use it with Claude Code, Codex, Cursor, and other coding agents, Vercel says to install the latest Vercel CLI and run npx vercel ai-gateway setup, then select humansand/ampersand-v0.1 in the agent.

Vercel describes AI Gateway as a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. The service also includes custom reporting, budgets for API keys, and routing rules. According to Vercel, AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference, including Bring Your Own Key requests.

For developers building shared chat experiences, the release provides a model-level mechanism for choosing between speaking and waiting, while the application remains responsible for executing and recording those actions. That separation gives teams control over delivery, persistence, and access checks as they integrate Ampersand into a conversation workflow.