CereBro

What is CereBro?

CereBro is a model-agnostic AI Agent Wrapper for .NET that implements the Model Context Protocol (MCP), allowing tools to be used with different AI models without changing code. It is built on the official C# SDK for MCP and currently supports OpenAI models.

How to use CereBro?

Install the NuGet packages Rob1997.CereBro and Rob1997.CereBro.Open-AI, create a servers.json file with MCP server configurations, set the OPEN_AI_API_KEY environment variable, and add the provided code to your Program.cs entry point. Build the host with BuildCereBro and call RunAsync to start the console‑based chat dispatcher.

Key features of CereBro

• Model‑agnostic AI agent wrapper for .NET
• Supports the Model Context Protocol (MCP)
• Currently works with OpenAI models
• Extensible via custom IChatDispatcher implementations
• Available as NuGet packages and a Unity package

Use cases of CereBro

• Building AI agents that can switch between different AI models without code changes
• Using MCP tools with OpenAI from .NET applications
• Creating custom chat interfaces with configurable tool integrations
• Integrating AI agent capabilities into Unity projects

FAQ from CereBro

Which AI models does CereBro support?

Currently only OpenAI models, with planned support for Claude, Grok, DeepSeek, Gemini, and Ollama.

Can I use multiple AI models at the same time?

No, at the moment CereBro does not support multiple models simultaneously.

How do I configure MCP servers for CereBro?

Create a servers.json file with an array of server objects specifying Id, Name, TransportType, and TransportOptions. The example uses stdio transport with npx.

What are the runtime requirements for CereBro?

A .NET environment, an OpenAI API key (for the supported provider), and the NuGet packages Rob1997.CereBro and Rob1997.CereBro.Open-AI.

Can I replace the built‑in chat dispatcher?

Yes, implement the IChatDispatcher interface and use the generic BuildCereBro<IChatDispatcher> overload.