SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI developer platformsupsonic

upsonic

Upsonic is a framework built in Python for autonomous AI agents, which can be used to create autonomous agents similar to OpenClaw and Claude Cowork, and also supports building traditional agent systems. It provides a collection of prebuilt agents, custom tool integration, MCP tool integration, and a unified OCR and document processing interface with a layered pipeline (supporting engines such as EasyOCR, RapidOCR, Tesseract, PaddleOCR, DeepSeek OCR, etc.), and can connect document indexing to coding tools such as Cursor. File and Shell operations are restricted to the workspace, and path traversal and dangerous commands are blocked. The project uses the MIT license.

Open website

About this tool

Upsonic is a Python-based framework for building autonomous AI agents. According to its repository description, it can be used to create autonomous agents similar to OpenClaw and Claude Cowork, and also supports building more traditional agent systems.

Purpose and Positioning This framework is mainly aimed at developers who want to build agents in Python, covering multiple forms from autonomous agents to traditional agent systems.

Main Capabilities

  1. Autonomous agents: Can create autonomous AI agents; all file and Shell operations are restricted to the workspace, and path traversal and dangerous commands are blocked.
  2. Prebuilt agents: Provides prebuilt autonomous agents built by the Upsonic community, each agent packaged with a skill, system prompt, and first message, allowing installation to running in just seconds; this collection is open to contributions, and you can submit a PR to add your own agent.
  3. Traditional agents and custom tools: Supports creating traditional agents and adding custom tools.
  4. MCP tool integration: Can connect MCP tools to connect agents to a large number of external data sources and services.
  5. OCR and document processing: Provides a unified OCR interface using a layered pipeline—Layer 0 handles document preparation (PDF to image, preprocessing), and Layer 1 runs the OCR engine. Supported engines include EasyOCR, RapidOCR, Tesseract, PaddleOCR, DeepSeek OCR, and DeepSeek used through Ollama.
  6. IDE integration: Upsonic documentation can be added as a source for coding tools, for example in Cursor through Settings → Indexing & Docs → Add https://docs.upsonic.ai/llms-full.txt to add; it also applies to VSCode, Windsurf, and similar tools.

How to Use

  • You can get started through the official documentation, Quickstart guide, and examples; the repository provides entry points such as Documentation, Quickstart, Examples, and Discord.
  • After creating an autonomous agent, you can further connect a Sandbox Provider (E2B) to obtain an isolated cloud execution environment.

Use Cases Suitable for development scenarios that need to build autonomous agents or traditional agent systems in Python, need to connect external tools and data sources (through MCP), or need document OCR processing capabilities; prebuilt agents are suitable for community users who want to quickly go from installation to running.

Limitations and Notes The repository materials do not provide specific installation commands, pricing and billing information, performance data, or platform runtime requirements; this content is not currently provided. When using file and Shell operations, they are restricted by the workspace, and path traversal and dangerous commands are blocked. The project is released under the MIT License, and specific details can be found in LICENCE (spelled as such in the original text); community contributions require first reading the Contributing Guide and Code of Conduct.

Related tools

Browse category