Add your expert

Add Your Expert to Fred#

Fred makes it easy to create custom AI agents (called “experts”) that respond to user questions, use tools, or collaborate in multi-step processes.

You can build:

• Simple agents that respond to user prompts
• Tool-using agents that call external services (e.g., via MCP)
• Complex agents that participate in a LangGraph-powered multi-agent workflow

Example: A Simple Expert#

You can create a new expert by subclassing AgentFlow. Here’s an example of a minimal generalist agent:

class GeneralistExpert(AgentFlow):
    name = "GeneralistExpert"
    role = "Generalist Expert"
    nickname = "Georges"
    ...

    async def async_init(self):
        self.model = get_model(self.agent_settings.model)
        self.base_prompt = "You are a helpful expert."
        self._graph = self._build_graph()
        super().__init__(...)

That’s it — the agent is ready to respond!

Tool-Using Agents#

Fred also supports agents that use tools via MCP servers. You can bind tools dynamically and build graphs that choose when to invoke them.

Static vs. Dynamic#

Agents can be registered statically (via config) or created dynamically at runtime via API. The same architecture applies in both cases.

Want to Go Deeper?#

See the Fred GitHub repository for:

• Design principles for agents
• Toolkits and MCP integration
• Examples: GeneralistExpert, DocumentsExpert, TabularExpert
• How to register and use agents

Start from the README – Agent Design section.

Ready to Build?#

Fred gives you full control with minimal boilerplate. Build fast. Deploy securely. Compose smart.