Studio

Visual UI for designing, deploying, and testing multi-agent systems built on Eigenoid Core.

Status	alpha
Version	0.1.0
Repo	eigenoid/app-studio
License	Business Source License 1.1

Overview

Eigenoid Studio is a guided builder that lets users design multi-agent systems without writing code. It is optimized for non-technical users first: start from a business goal, shape agents in plain language, choose AI providers and models, define skills as natural-language instructions, and deploy the system from the same UI.

An advanced architecture canvas is available for power users who want to work directly with the system model, node topology, and delegation configuration.

What you can do

• Describe what your AI team should accomplish in business language
• Define agents with friendly names, responsibilities, and provider/model selection (OpenAI, Google Gemini, Anthropic, OpenRouter)
• Author skills using natural language instead of code
• Connect agents with simple delegation handoffs
• Deploy the system and test it with a built-in chat interface
• Inspect the generated eigenoid.yaml and Python files
• Drop into the advanced canvas for full topology control

Architecture

Tech stack

Layer	Technology
UI framework	React 18, TypeScript
State management	Zustand
Build tool	Vite 5
Visual architecture	React Flow (xyflow)
YAML generation	js-yaml
Styling	Tailwind CSS 3, PostCSS

Setup

Frontend only

cd app-studio
npm install
npm run dev

bash

Open http://localhost:5173.

Full stack (guided builder + deploy + chat)

Requires Eigenoid Core checked out locally:

# Clone core alongside studio
git clone https://github.com/eigenoid/eigenoid.git ../eigenoid

# Or point to an existing checkout
export EIGENOID_CORE_ROOT=/path/to/eigenoid

# Start backend + frontend
bash start.sh

bash

Build for production

npm run build

bash

Output goes to dist/.

Guided flow

The five-step guided flow is designed so that a user with no engineering background can build and deploy a working multi-agent system:

1. Plan -- describe the outcome in business language, pick a starter team or blueprint.
2. Agents -- define each agent with a name, responsibility, AI provider/model, and natural-language skills.
3. Flow -- connect agents with delegation handoffs. The advanced canvas is available for complex topologies.
4. Launch -- deploy the system, view logs, and test with the built-in chat interface.
5. Outputs -- inspect the generated eigenoid.yaml and Python files.

AI provider configuration

Studio supports per-agent model selection. Each agent references an environment variable for its API key:

Provider	Env var
OpenAI	OPENAI_API_KEY
Google Gemini	GEMINI_API_KEY
Anthropic	ANTHROPIC_API_KEY
OpenRouter	OPENROUTER_API_KEY

The generated YAML includes model, api_key, and optionally api_base_url fields for each agent.