Upsonic/Upsonic

Project overview

A Python framework for developers who need to build autonomous or traditional AI agents with custom tools, alongside a unified OCR pipeline. It is installable via pip and restricts agent file and shell operations to a designated workspace.

Project type

AI Agent · Document Processing

Use cases

Documents & Office · Automation

Deployment

Refer to project documentation

License

MIT

Best for

• Python developers and AI engineers who need a programmable library to build or extend AI agents with custom tools.
• Teams that require document OCR processing alongside agent task execution.

Key capabilities

• Enables building autonomous AI agents that can perform tasks like analyzing server logs within a restricted workspace.
• Supports building traditional agent systems for tasks like market trend analysis.
• Allows adding custom tools to agents using a decorator pattern.
• Provides a unified OCR interface with a layered pipeline: Layer 0 handles document preparation (PDF to image conversion, preprocessing), Layer 1 runs the OCR engine.
• Ready-to-run agents built by the Upsonic community, each packaging a skill, system prompt, and first message so you can go from install to running in seconds.

Limitations and risks

• File and shell operations executed by agents are restricted to a designated workspace.
• Autonomous agents may execute shell commands; safety blocks exist but the capability remains.

Getting started

• Install the framework easily using uv pip install upsonic to begin defining agents and tasks.

Evidence and sources

• GitHub project description: Build autonomous AI agents in Python.
• README: **Build Autonomous AI Agents in Python**
• README: Upsonic is a Python framework for building autonomous agents like OpenClaw and Claude Cowork, as well as more traditional agent systems.
• README: [](https://badge.fury.io/py/upsonic)
• README: [](https://docs.upsonic.ai)

Reviewed at: 2026-07-18T06:14:20Z