Why Fred - A Developer Manifesto

You want to build an agent that does more than answer questions.

You want it to:

• Use tools
• Coordinate with others
• Run in production
• And not fight with lifecycle, async bugs, or YAML spaghetti.

Fred is for you.#

Fred is an agentic framework built by practitioners, not just to demo ideas — but to run real agents, securely and repeatedly.

🔍 What Fred gives you (that’s surprisingly rare)#

• ✅ LangGraph-powered agents, with multi-step flows, conditionals, and state.
• ✅ Tools + Experts — like TabularExpert, DocumentsExpert, RagsExpert — that actually work out of the box.
• ✅ Leader agents that plan, delegate, and summarize results across experts.
• ✅ Built-in support for OpenAI, Azure, Ollama, and real backend integration (RAG, vector search, SQL).
• ✅ Full lifecycle support: from dev notebooks to Docker to Kubernetes.

🚀 Agents are easy to build#

Define a new agent in seconds:

class MyAgent(AgentFlow):
    def __init__(self, settings: AgentSettings):
        self.model = get_model(settings.model)
        self._graph = self._build_graph()
        super().__init__(..., model=self.model, graph=self._graph)

    def _build_graph(self): ...

Need tools? Add toolkit, bind tools, and use ToolNode.
Need async? Just implement async_init() — it’s all wired for you.

Fred abstracts the hard parts — async FastAPI, stateful memory, WebSocket streaming — so you can focus on logic.

🧩 Static or Dynamic#

Fred supports:

• ✅ Declarative YAML-based agent loading
• ✅ Dynamic runtime creation (via API or code)
• ✅ Integration with MCP services (toolchains, databases, clusters, etc.)

Everything is typed, clean, and minimal.

What takes time without Fred#

• Getting async + tool + graph orchestration to not break
• Reusing models and toolkits across sessions without bugs
• Hooking LangGraph + WebSockets + FastAPI the right way
• Explaining to others what’s happening

You’ll spend hours or days debugging what Fred gives you in minutes.

For data scientists too#

Want to test your toolchain in a notebook? Easy.

agent = TabularExpert(agent_settings)
await agent.async_init()
await agent.compiled_graph.ainvoke({"messages": [HumanMessage(...)]})

Fred works with LangChain and LangGraph directly.
No lock-in. No code generation. Just Python.

TL;DR#

Fred is the missing layer between LangChain experiments and real agentic applications.
It’s built to help you create serious, tool-using, multi-agent systems without getting lost in infra.

Try it.