Upsonic is an AI agent development framework used by fintech leaders and tested at their scale against attacks and reasoning puzzles.

pip install upsonic

from upsonic import Task, Agent

task = Task("Who developed you?")

agent = Agent(name="Coder")

agent.print_do(task)

See our guides to jumpstart your AI agent within minutes. We design them to onboard the new users to the framework.

1. Create an Agent
2. Create a Task
3. Add a Safety Engine
4. Add a Tool
5. Add an MCP
6. Integrate a Memory
7. Creating a Team of Agents

Upsonic provides a feature set to build safety-first, high-performance AI Agents. It helps you go to production without spending hours on research and boilerplate. These are the main parts:

• Safety First: Upsonic provides its own Safety Engine that manages User and Agent messages and checks their status for your policies. You can customize it by designing new rule and action sets.
• Direct LLM Calls: In Upsonic we support the same interface for your whole AI operations. You don't need to go with another framework to complete your small jobs.
• Structured Outputs: Upsonic sets agent outputs to make them Python objects. So you can integrate your application without struggling with LLM outputs.
• Built-in RAG and Memory: In Upsonic you can create world class . We support the Agentic RAG, Memory Logics and providers of them.
• Customizable Memory Logics: You are able to create memories that focus on user, event and chat. Also you are free to use Local and Cloud databases.
• Agent Teams: Upsonic provides the most reliable agent team architecture with memory, context management and leader agent.
• FastAPI Compatible Agents: You can turn your agents into production-ready APIs
• Tracking the Executions: You can use Upsonic AgentOS to get the execution history, monthly costs andresponse times of your agents.
• Deploy at scale: Upsonic agents work in the greatest and fastest-growing fintech companies and scaling is available on Upsonic AgentOS.

You can access our documentation at docs.upsonic.ai All concepts and examples are available there.

Upsonic uses anonymous telemetry to help us understand how the framework is being used and improve our development focus. We are committed to transparency and user privacy.

All telemetry is anonymous - we only track a randomly generated system ID to distinguish unique installations.

Telemetry helps us:

• 🎯 Focus development on frequently-used features
• 🐛 Identify and fix common errors and edge cases
• 📈 Understand performance characteristics at scale
• 🔧 Improve framework reliability

You can completely disable telemetry in multiple ways:

Option 1: Environment Variable (Recommended)

export UPSONIC_TELEMETRY=false

Option 2: In Python Code

import os
os.environ["UPSONIC_TELEMETRY"] = "false"

from upsonic import Agent  # Import after setting env var

Option 3: .env File

# .env
UPSONIC_TELEMETRY=false

Once disabled, no data will be sent to our telemetry service.