open multi agent canvas

Report Abuse

Basic Information

Open Multi-Agent Canvas is an open-source web application that provides a unified chat interface to run and manage multiple AI agents within a single dynamic conversation. The project combines a Next.js frontend with CopilotKit and LangGraph integrations to orchestrate agent interactions and supports common use cases such as travel planning, research, and general-purpose multi-step tasks. It includes a built-in MCP (Model Context Protocol) agent and references example agents that can be run separately or deployed on LangSmith. The repository contains frontend code and an optional agent backend, instructions for local development, and configuration for connecting to external MCP-compatible servers. It is intended for users who want a multi-agent UI and for developers who want to connect or extend agents via MCP servers and local processes.

Links

Categorization

App Details

Features
Multi-agent chat canvas that lets you add and manage multiple agents in a single conversation. Built with Next.js for the frontend and integrated with CopilotKit and LangGraph for agent orchestration. Includes a built-in MCP Agent able to connect to MCP-compatible servers and a built-in math server. Support for configuring custom MCP servers via a GUI with Standard IO (local scripts) and SSE endpoints. Examples and prebuilt agents for travel planning and AI research are provided in separate repositories. Deployment and development instructions include environment variables for API keys, pnpm-based frontend workflow, an optional backend that uses poetry and langgraph for local agent hosting, and guidance for running tunnels for local development.
Use Cases
This repository helps users and developers coordinate multiple AI agents from a single interface, simplifying complex workflows like travel planning and research by letting agents collaborate in one conversation. Developers can extend the system by adding MCP-compatible backends, connecting external SSE servers, or running local scripts via Standard IO to incorporate custom logic or services. The included examples and the optional backend make it easier to test, run, and deploy agent configurations, and the project provides concrete setup steps, required environment variables, and tooling suggestions to get a local instance running. It is useful for teams building multi-agent prototypes, experimenting with agent orchestration, or deploying multi-channel agent flows.

Please fill the required fields*