Skip to content

flo7up/cronosaurus

BranchesTags

Repository files navigation

Cronosaurus

Cronosaurus

Open-source multi-agent AI platform
Azure AI Foundry • OpenAI • Anthropic

License Python Node

Create autonomous AI agents with a rich tool ecosystem, schedule recurring tasks, connect email accounts, and extend capabilities through MCP servers — all from a sleek chat interface.

Features

  • Multi-agent — Spin up as many agents as you need, each with its own tools, model, and conversation history
  • Multi-provider — Use Azure AI Foundry, OpenAI, or Anthropic as your LLM backend
  • Tool ecosystem — Built-in tools for crypto, stocks, web search, email, Azure cost analysis, prediction markets, weather, Twitch stream capture, and more
  • Agent collaboration — Agents can discover, inspect, and message each other for multi-agent workflows
  • Triggers — Schedule agents to run tasks on a recurring basis (e.g. "email me a market summary every morning")
  • Email integration — Send and read email via SMTP/IMAP, with Gmail push notification support
  • MCP servers — Extend agents with any Model Context Protocol server
  • Vision — Paste images or capture Twitch stream thumbnails for visual analysis
  • Notifications — In-app bell + optional email alerts so agents can proactively reach out
  • Todo lists — Agents break complex tasks into visible, trackable step-by-step lists
  • Onboarding wizard — Guided first-run setup — no .env editing required

Screenshots

Azure Cost Analysis Building Progress Monitoring Mortgage Rate Tracking
Azure Cost Analysis Building Progress Monitoring Mortgage Rate Tracking
Tool Management Self Tool Activation Agent-to-Agent Collaboration
Tool Management Self Tool Activation Agent-to-Agent Collaboration

Demo Video

Open the demo video directly

Architecture

┌─────────────────┐        ┌──────────────────────┐
│   React + TS    │  HTTP  │   FastAPI (Python)    │
│   Vite + TW     │◄──────►│                       │
│   Frontend      │  SSE   │   Routers / Services  │
└─────────────────┘        └───────┬──────┬────────┘
                                   │      │
                        ┌──────────┘      └──────────┐
                        ▼                            ▼
              ┌──────────────────┐        ┌──────────────────┐
              │  LLM Provider    │        │  Azure Cosmos DB  │
              │  (Foundry /      │        │  (State + Data)   │
              │   OpenAI /       │        └──────────────────┘
              │   Anthropic)     │
              └──────────────────┘
Layer Tech
Frontend React 19, TypeScript, Vite, Tailwind CSS
Backend Python 3.12, FastAPI, Uvicorn
AI Azure AI Foundry, OpenAI, or Anthropic (configurable)
Database Azure Cosmos DB (NoSQL)

Model Providers

Set MODEL_PROVIDER in your .env (or use the onboarding wizard) to choose one:

Provider MODEL_PROVIDER What you need Conversation history
Azure AI Foundry azure_foundry Azure subscription + AI Foundry project Stored server-side in Azure Agent Service
OpenAI openai OpenAI API key Stored in Cosmos DB
Anthropic anthropic Anthropic API key Stored in Cosmos DB

Note: All providers require Azure Cosmos DB for agent definitions, user settings, tool configs, and email accounts.

Built-in Tools

Tool Description
Crypto Live prices and order book depth from Hyperliquid DEX
Stocks Real-time prices, charts, and fundamentals from Yahoo Finance
Send Email Compose and send emails via SMTP
Read Email Read and search emails via IMAP
Web Search Search the web and scrape pages via DuckDuckGo
Polymarket Prediction market odds, trending events, and search
Azure Costs Spending breakdown by resource group, service, or resource
Weather Current conditions and 7-day forecasts (Open-Meteo)
Twitch Capture Screenshot live Twitch streams for visual analysis
Agent Collaboration Agents can list, inspect, and message other agents
Triggers Schedule recurring automated tasks
Notifications In-app bell and/or email alerts
Todo Lists Break complex tasks into trackable steps
Confirmations Interactive approve/reject buttons for sensitive actions

Tools are modular — enable only what each agent needs. Custom tools can be added by dropping a file into backend/app/tools/custom/.

Quick Start

Prerequisites

  • Python 3.12+
  • Node.js 20+ and npm
  • Azure Cosmos DB NoSQL account
  • One LLM provider: Azure AI Foundry, OpenAI, or Anthropic

1. Clone the repository

git clone https://github.com/<your-username>/cronosaurus.git
cd cronosaurus

2. Start the application

Option A: PowerShell launcher (Windows)

.\start.ps1          # Opens backend + frontend in separate terminals

Option B: npm (cross-platform)

npm run dev           # Runs backend + frontend concurrently

Option C: Manual setup

Backend:

cd backend
python -m venv venv
# Windows: venv\Scripts\activate
# macOS/Linux: source venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Frontend:

cd frontend
npm install
npm run dev

3. Complete the onboarding wizard

Open http://localhost:5173 — a guided wizard will walk you through:

  1. LLM Provider — Choose Azure AI Foundry, OpenAI, or Anthropic and enter credentials
  2. Models — Select which models appear in the model selector
  3. Cosmos DB — Provide your database URL and key
  4. Tools — Optionally enable email and other integrations

All settings are saved to backend/settings.json and can be changed anytime from Management Panel > Settings.

Already have a .env file? If backend/.env has your provider settings and COSMOS_URL + COSMOS_KEY, onboarding is skipped automatically.

Authentication

Azure AI Foundry (Keyless)

Authenticates via DefaultAzureCredential — no API keys needed. Just az login.

Required roles on your AI Foundry project:

Role Purpose
Azure AI Developer Invoke models, create and manage agents
Azure AI Inference Deployment Operator List deployments (for "Load from Foundry" feature) 
USER_ID=$(az ad signed-in-user show --query id -o tsv)
RESOURCE_ID="/subscriptions/<sub>/resourceGroups/<rg>/providers/Microsoft.MachineLearningServices/workspaces/<project>"

az role assignment create --role "Azure AI Developer" --assignee "$USER_ID" --scope "$RESOURCE_ID"
az role assignment create --role "Azure AI Inference Deployment Operator" --assignee "$USER_ID" --scope "$RESOURCE_ID"

OpenAI

MODEL_PROVIDER=openai
OPENAI_API_KEY=sk-...
OPENAI_MODEL=gpt-4.1-mini

Anthropic

MODEL_PROVIDER=anthropic
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL=claude-sonnet-4-20250514

Configuration Reference

All settings via environment variables or backend/.env:

General

Variable Required Default Description
MODEL_PROVIDER No azure_foundry azure_foundry, openai, or anthropic
FRONTEND_URL No http://localhost:5173 Allowed CORS origin
PORT No 8000 Backend listen port
LOG_LEVEL No INFO DEBUG, INFO, WARNING, ERROR

Azure AI Foundry

Variable Required Default Description
PROJECT_ENDPOINT Yes — AI Foundry project endpoint
MODEL_DEPLOYMENT_NAME No gpt-4o Default model deployment

OpenAI

Variable Required Default Description
OPENAI_API_KEY Yes — OpenAI API key
OPENAI_MODEL No gpt-4.1-mini Default model

Anthropic

Variable Required Default Description
ANTHROPIC_API_KEY Yes — Anthropic API key
ANTHROPIC_MODEL No claude-sonnet-4-20250514 Default model

Azure Cosmos DB (required)

Variable Required Default Description
COSMOS_URL Yes — Cosmos DB account URL
COSMOS_KEY Yes — Cosmos DB primary key
COSMOS_DB No cronosaurus Database name
COSMOS_CONNECTION_STRING No — Alternative to URL + key

Other

Variable Required Default Description
EMAIL_ENCRYPTION_KEY No — Encryption key for SMTP passwords at rest (falls back to COSMOS_KEY)

Extending Cronosaurus

Adding MCP Servers

  1. Click the tools icon in the agent header
  2. Click "Add more tools…" → MCP tab
  3. Add your MCP server URL (and optional API key)
  4. Tools appear automatically in the agent's tool dropdown

Adding Custom Tools

Drop a Python file into backend/app/tools/custom/ following the template. It will be auto-discovered on startup and appear in the Tool Library.

Setting Up Email

  1. Management Panel → Email tab
  2. Enter SMTP/IMAP server details and credentials
  3. Passwords are encrypted at rest with Fernet (AES-128-CBC + HMAC-SHA256)
  4. Enable Send Email / Read Email tools on your agent

For Gmail, use an App Password with IMAP enabled.

Project Structure

cronosaurus/
├── backend/
│   ├── app/
│   │   ├── main.py              # FastAPI app + lifespan
│   │   ├── config.py            # Settings via pydantic-settings
│   │   ├── models/              # Pydantic request/response models
│   │   ├── routers/             # API route handlers
│   │   ├── services/            # Business logic (agent service, store, scheduler)
│   │   │   └── providers/       # LLM provider implementations
│   │   └── tools/               # Tool implementations
│   │       └── custom/          # Drop-in custom tools (auto-discovered)
│   ├── requirements.txt
│   └── Dockerfile
├── frontend/
│   ├── src/
│   │   ├── App.tsx              # Main app shell
│   │   ├── components/          # React components
│   │   ├── api/                 # API client functions
│   │   └── types/               # TypeScript types
│   ├── package.json
│   └── Dockerfile
├── start.ps1                    # PowerShell launcher (Windows)
└── package.json                 # Root dev script (concurrently)

Docker

# Backend
cd backend
docker build -t cronosaurus-backend .
docker run -p 8000:8000 --env-file .env cronosaurus-backend

# Frontend
cd frontend
docker build -t cronosaurus-frontend .
docker run -p 80:80 cronosaurus-frontend

Security

  • Secrets loaded from environment variables — no credentials hardcoded
  • SMTP passwords encrypted at rest with Fernet (AES-128-CBC + HMAC-SHA256)
  • CORS locked to configured FRONTEND_URL
  • Azure AI Foundry uses DefaultAzureCredential (keyless)
  • OpenAI / Anthropic keys stored only in .env

Contributing

Cronosaurus is designed to be extended. See CONTRIBUTING.md for guides on adding custom tools and triggers.

License

MIT

About

Open-source multi-agent AI platform with scheduled triggers, tool ecosystem, and support for different llm providers

Topics

microsoft opensource scheduling cronjob cronjob-scheduler workflow-automation ai-agents agentic-workflow scheduler-agent

Resources

Readme

License

MIT license

Code of conduct

Code of conduct

Contributing

Contributing

Security policy

Security policy
Activity

Stars

3 stars

Watchers

1 watching

Forks

2 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages