Fred

Value Proposition

Agentic innovation meets production-grade engineering

Fred is both:

  • An innovation lab — to help developers rapidly explore agentic patterns, domain-specific logic, and custom tools.
  • A production-ready platform — already integrated with real enterprise constraints: auth, security, document lifecycle, and deployment best practices.

Why it’s different

Fred and knowledge-flow are more than just a wrapper around an LLM or a UI for chatting. They’re an end-to-end framework for building intelligent, grounded, and extensible agents — at scale.

Full agentic backend

Fred uses LangGraph to model complex workflows across multiple agents:

  • Delegation to domain-specific agents (Dominic, Tessa, Rico, etc.)
  • Streaming responses and tool integration
  • WebSocket + REST APIs for real-time interactivity
  • Agents subclass AgentFlow, with clean async_init() lifecycles

Tool-augmented experts

Fred makes tool-using agents a first-class citizen. Each expert can bind to:

  • External tools via MCP (e.g., SQL over tabular data, vector search)
  • Custom LangChain-compatible tools
  • Prebuilt LangGraph nodes (e.g., ToolNode) for seamless execution

Agents like TabularExpert and RagsExpert show how external capabilities are integrated with reasoning and planning — all declaratively.

Document-aware reasoning

The knowledge-flow backend handles ingestion, parsing, chunking, vectorization, and retrieval:

  • Structured + unstructured support (PDFs, Markdown, CSV, SQL)
  • UID-based access and versioning
  • Seamless agent integration (e.g., tool messages directly grounded in source content)

Modular, composable, and decoupled

Fred and knowledge-flow communicate via clean APIs and support MCP (Multi-Component Protocol) to connect to:

  • External toolchains (like a Kubernetes analyzer)
  • Custom vector or tabular backends
  • Remote agents, services, or pipelines

You can plug in new components — or replace existing ones — without rewriting the system.

Built for real deployment

Fred is production-minded:

  • Authentication and access control
  • Configurable via YAML + .env
  • On-prem and cloud-ready
  • Clean React UI backed by RTK Query and WebSockets
  • Compatible with Docker, Dev Containers, Kubernetes

Compared to typical OSS agent frameworks

Feature / GoalFred + knowledge-flowOpenWebUI / Flowise / etc.
Multi-agent orchestration✅ LangGraph workflows❌ Often single-agent logic
Secure document lifecycle✅ Full content + metadata mgmt❌ Basic file upload only
Tool-augmented experts✅ MCP + bindable LangChain tools❌ Minimal or static tools
Async agent lifecycle✅ async_init() with tools + graph❌ Often blocking / static
Modular components (MCP)✅ Plug in custom tools/agents❌ Tooling often hardcoded
Real-time streaming & UI✅ WebSocket + RTK Query⚠️ Often polling or static
Co-innovation and governance✅ Designed for partnerships❌ Lacks clear extension model

Open Source by Intent

Fred is open source not to reinvent frameworks, but to:

  • Enable co-innovation with partners and contributors
  • Make enterprise-grade agentic development accessible
  • Provide a foundation that bridges open tooling with secure, extensible infrastructure

Learn More

Summary

Fred and knowledge-flow aren’t just another chatbot wrapper.
They are a foundation for building intelligent, secure, and domain-aware LLM applications, backed by strong engineering, modular design, and a focus on real-world usability.

Whether you want to build a Kubernetes explainer, a financial analyst, or a private document assistant — Fred gives you the pieces to do it right.

Prev
User Interface
Next
Resources