Happy slice of mcp pizzaMCP.pizza

open-agent-platform

MCP.Pizza Chef: langchain-ai

This project offers a visual web dashboard where anyone can create, configure, and chat with custom AI agents, connect those agents to outside tools through the Model Context Protocol, and even have several agents collaborate on one task. Using it requires setting up hosting for your agents and supplying an API key for an AI model such as OpenAI, Claude, or Gemini. Note that the creators have discontinued this project and now point new users toward a hosted alternative called Agent Builder.

Archived · The author made this repository read-only on GitHub.
Coding
Data

Use This MCP client To

Build a custom AI agent without writing any code Connect an agent to my team's tools and files Have several agents work together on one task Search my own documents for more accurate answers Control exactly who can access and use each agent Customize the questions and settings my agent supports

README

Open Agent Platform

Caution

This repository has been deprecated.

We recommend using Agent Builder on LangSmith instead. Agent Builder provides a fully managed, no-code experience for creating, testing, and deploying agents — no self-hosting required.

Original README (archived)

Open Agent Platform is a no-code agent building platform. These agents can be connected to a wide range of tools, RAG servers, and even other agents through an Agent Supervisor!

oap-simple-demo.mp4

Overview

Open Agent Platform provides a modern, web-based interface for creating, managing, and interacting with LangGraph agents. It's designed with simplicity in mind, making it accessible to users without technical expertise, while still offering advanced capabilities for developers.

Key Features

  • Agent Management: Build, configure, and interact with agents through an intuitive interface.
  • RAG Integration: First-class support for Retrieval Augmented Generation with LangConnect.
  • MCP Tools: Connect your agents to external tools through MCP servers.
  • Agent Supervision: Orchestrate multiple agents working together through an Agent Supervisor.
  • Authentication: Built-in authentication and access control.
  • Configurable Agents: Easily define how users can configure your agents through a rich UI.

Documentation

For detailed setup instructions, guides, and API references, please see the docs directory.

Getting Started

To quickly get started with Open Agent Platform, check out the Quickstart Guide in our documentation.

Community and Support

We encourage you to explore the platform, build your own agents, and contribute to the Open Agent Platform community!

Concepts/FAQ

Concepts

Agents

An agent is a custom configuration on-top of an existing LangGraph graph. This is the same concept as an assistant, in the LangGraph API.

FAQ

Do I need a backend server?

OAP does not require a standalone backend server to be running in order for the web app to work. As long as you've added deployments from LangGraph Platform to your instance, it should work as expected!

However, if you want to use the RAG features, you will need to have the LangConnect server running on its own. This is because the LangConnect RAG server is intended to be hosted independently from your LangGraph deployments. See the LangConnect docs for more information.

How can I build my own agents?

Yes! See the Building Your Own Agents section for more information.

How can I use non-supabase auth?

Yes! It requires some modifications to be made to the code, but we've implemented authentication in a way which makes it easy to swap out with any other authentication provider.

How can I use non-langgraph agents?

No. All agents you intend to use with OAP must be LangGraph agents, deployed on LangGraph Platform.

Why is my agent's config is only showing string inputs, and not custom fields?

First, ensure you're using the latest version of LangGraph. If running locally, make sure you're using the latest version of the LangGraph API, and CLI packages. If deploying, make sure you've published a revision after 05/14/2025. Then, check that you have the x_oap_ui_config metadata set on your configurable fields. If you have, check that your configurable object is defined using LangGraph Zod (if using TypeScript), as this is required for the Open Agent Platform to recognize & render your UI fields.

If it's still not working, confirm your x_oap_ui_config metadata has the proper fields set.

open-agent-platform FAQ

Can I use this to build a customer-support chatbot without any coding?
Yes — you can create and configure an agent through the visual interface, connect it to your own documents for answers, and publish it for others to use, all without writing code.
Can I connect this to my own files so my agent gives more accurate answers?
Yes — the built-in retrieval feature, called RAG, lets your agent search through documents you upload before it answers a question.
Is this still being actively developed and supported?
No — the creators have discontinued Open Agent Platform and now recommend Agent Builder, a newer hosted alternative on LangSmith, for anyone getting started today.
Does this work inside apps I already use, like Claude Desktop or ChatGPT?
No — Open Agent Platform is a separate, self-hosted web application rather than a plug-in for existing chat apps such as Claude Desktop, ChatGPT, or Gemini.
Do I need an account or an API key to use it?
Yes — you'll need a hosting setup for your agents (a LangGraph deployment) and an API key for whichever AI model you choose to power them, such as OpenAI, Claude, or Gemini.
How difficult is it to set up?
It is best suited to someone comfortable with technical setup, since it requires deploying and hosting the platform yourself rather than clicking a single install button.
Can multiple agents work together on the same task?
Yes — a feature called Agent Supervisor lets you connect several agents so they can coordinate and hand off work to one another.
Can I limit who is allowed to see or use an agent I build?
Yes — built-in authentication and access controls let you decide who can view, configure, or run each agent.