Neuro SAN Studio

Your launchpad for building intelligent multi-agent systems. Neuro SAN Studio is a hands-on playground for the Neuro SAN framework, featuring ready-to-run examples, tutorials, and tools that let you design, test, and deploy sophisticated agent networks in minutes—not months. Whether you're a researcher exploring adaptive AI systems, a developer prototyping production solutions, or a domain expert configuring agents without code, this studio handles the orchestration complexity so you can focus on solving real problems.

---

Neuro SAN is the open-source library powering the Cognizant Neuro® AI Multi-Agent Accelerator, allowing domain experts, researchers and developers to immediately start prototyping and building agent networks across any industry vertical.

---

Neuro SAN library

What is Neuro SAN?

Neuro AI system of agent networks (Neuro SAN) is an open-source, data-driven multi-agent orchestration framework designed to simplify and accelerate the development of collaborative AI systems. It allows users—from machine learning engineers to business domain experts—to quickly build sophisticated multi-agent applications without extensive coding, using declarative configuration files (in HOCON format).

Neuro SAN enables multiple large language model (LLM)-powered agents to collaboratively solve complex tasks, dynamically delegating subtasks through adaptive inter-agent communication protocols. This approach addresses the limitations inherent to single-agent systems, where no single model has all the expertise or context necessary for multifaceted problems.

---

✨ Key Features

• 🗂️ Data-Driven Configuration: Entire agent networks are defined declaratively via simple HOCON files, empowering technical and non-technical stakeholders to design agent interactions intuitively.
• 🔀 Adaptive Communication (AAOSA Protocol): Agents autonomously determine how to delegate tasks, making interactions fluid and dynamic with decentralized decision-making.
• 🔒 Sly-Data: Sly Data facilitates safe handling and transfer of sensitive data between agents without exposing it directly to any language models.
• 🧩 Dynamic Agent Network Designer: Includes a meta-agent called the Agent Network Designer – essentially, an agent that creates other agent networks. Provided as an example with Neuro SAN, it can take a high-level description of a use-case as input and generate a new custom agent network for it.
• 🛠️ Flexible Tool Integration: Integrate custom Python-based "coded tools," APIs, databases, and even external agent ecosystems (Agentforce, Agentspace, CrewAI, MCP, A2A agents, LangChain tools and more) seamlessly into your agent workflows.
• 📈 Robust Traceability: Detailed logging, tracing, and session-level metrics enhance transparency, debugging, and operational monitoring.
• 🌐 Extensible and Cloud-Agnostic: Compatible with a wide variety of LLM providers (OpenAI, Anthropic, Azure, Ollama, etc.) and deployable in diverse environments (local machines, containers, or cloud infrastructures).

---

Use Cases

Here are a few examples of use-cases that have been implemented with Neuro SAN. For more examples, check out docs/examples.md.

And many more: check out docs/examples.md.

---

High level Architecture

---

Install

These instructions are for Linux and macOS systems. Please adjust the commands accordingly for Windows.

Install uv

uv is a fast Python package and project manager built by Astral.

Official installation docs: 👉 https://docs.astral.sh/uv/getting-started/installation/

Create a new Python project

Create a folder for your project:

mkdir my_project
cd my_project

Create a virtual environment, initialize a git repo and install neuro-san-studio

uv init
uv venv
source .venv/bin/activate
uv add neuro-san-studio

Initialize neuro-san-studio

Run ns init to initialize a Neuro SAN Studio project. ns stands for Neuro SAN. You can also use the long command neuro-san-studio instead. It will:

• let you choose an LLM provider
• create a config folder with your choice of LLM models and plugins configuration
• create an mcp folder with a list of MCP tools
• create a registries folder with a simple agent network and the Agent Network Designer, so you can start designing your own networks right away
• create coded_tools and middleware folders with the Python code those agent networks need

To learn more about the ns command run ns --help.

ns init

Which LLM providers do you want to enable?

#  Provider       Default model
1  OpenAI         gpt-5.2 (default)
2  Anthropic      claude-sonnet
3  Google Gemini  gemini-3-flash

Enter numbers separated by commas (default: 1):

Note: To access all Neuro SAN Studio capabilities, clone this repository and follow the setup instructions in the docs/dev_guide.md.

Set your LLM API key(s)

1. Set your provider key, e.g. OPENAI_API_KEY, ANTHROPIC_API_KEY or GOOGLE_API_KEY (or create a .env file in the current directory). See docs/api_key.md for details and other providers.

   export OPENAI_API_KEY="XXX"
   

2. Check your LLM API keys are correctly configured:

   ns check-llm-keys
   

3. Check your config/llm_config.hocon is working:

   ns check-config
   

   If the configuration is valid you will get a hello response from the configured LLMs.

Import agent networks

ns init already installs the Agent Network Designer, so you can start building right away. Use ns import to add any of the other agent network examples that ship with neuro-san-studio.

Run it with no arguments to pick from an interactive list:

ns import

Or name a group or a single network directly:

ns import basic         # every network in the "basic" group
ns import hello_world   # a single network

Each imported network brings its dependencies with it -- coded tools, middleware, sub-networks -- and is registered in registries/manifest.hocon. A running server picks it up within a few seconds.

See docs/cli/import.md for the full set of options.

Start the developer UI

You can start a neuro-san server and the nsflow UI with the ns run command:

ns run

The Neuro SAN server listens on localhost:8080.

The nsflow UI is served at http://localhost:4173/.

Logs land under logs/ (server.log, nsflow.log, thinking_dir/).

Screenshot:

Agent Network Designer

Use the Agent Network Designer to create your own agent network.

1. From the nsflow UI, click the NEW button at the top, center of the screen.
2. In the new window that opens, type your prompts in the text box in the bottom right corner of the screen. Then Agent Network Designer:
   • Creates the agents
   • Links them together
   • Writes instructions for each agent
   • Generates a few sample queries you can ask this agent network
   • Saves the agent network in the registries/generated folder
3. Once the Agent Network Designer is done and comes back with an answer in the chat window, you can continue the design by asking it to make changes
4. Once you're happy with the design, test it! Click the blue Launch button at the top center of the screen. It opens a new window from which you can chat with the agent network.
5. If you want to make modifications, go back to the editor window and ask for changes.
6. You can also edit any agent network by clicking the pen icon next to its name in the main window.

Import a project from a file / Export to a file

You can import a project from a .hocon file or from a zip file using the ns import <PATH>.

ns import ~/Downloads/my_project.hocon

Similarly, you can export an agent network and all its dependencies using the ns export command:

ns export my_project.hocon

See docs/cli/export.md for details.

Command reference

Use ns <command> --help for the full flag list of any subcommand.

---

User guide

Ready to dive in? Check out the user guide for a detailed overview of the neuro-san library and its features.

---

Tutorial

For a detailed tutorial, refer to docs/tutorial.md.

---

Examples

For examples of agent networks, check out docs/examples.md.

---

Developer Guide

For local development setup and contribution instructions, see the docs/dev_guide.md.

---

Community Projects

Applications

• Climate Change: a tool to answer questions about COP, the Paris Agreement or the Kyoto Protocol using UNFCCC documents.
• Enterprise Access Portal: an AI-powered multi-agent system for managing enterprise application access requests and IT operations.
• F1 fans eval: an app that evaluates F1 fan submissions about why they are the biggest F1 fans.
• PDF Knowledge Assistant: a Flask web app that queries PDFs using RAG with topic-based long-term memory synthesis across documents.
• Loopy Agents: run Neuro SAN agents continuously or on triggers through a separate service, with asynchronous messaging.
• Annual Report Reader: analyzes a LinkedIn profile and delivers a personalized summary of Cognizant's 2024 Annual Report, surfacing content most relevant to the user's industry and seniority level.
• Tochiro File Organizer: a macOS file organization assistant with a dedicated UI to analyze a folder, create a plan for moving the files, ask for approval and execute the moves.
• Legacy Business-Rule Extractor: a 6-agent network that extracts business rules from legacy COBOL, Java, and PL/SQL code, pairing deterministic CodedTool parsers with LLM agents to produce a modernization-ready specification document.

Utilities

• Neuro SAN Web Client: a basic Flask web client interface for Neuro SAN.
• Neuro SAN Slack app a Slack integration that lets you interact with Neuro SAN directly from your workspace.

---

Links

• Website: Cognizant AI Lab
• YouTube: Decision AI
• X: @cognizantailab
• LinkedIn: Cognizant AI Lab
• Amazon Marketplace: Cognizant Neuro SAN
• Azure Marketplace: Cognizant Neuro SAN

---

More details

For more information, check out the Cognizant AI Lab Neuro SAN landing page.