A compact curated collection of open-source Python and AI tools useful for building modern AI applications, agents, document pipelines, and RAG systems.
A compact curated list of document parsing, OCR, PDF extraction and document ETL tools for AI/RAG pipelines.
- Docling — Structured document conversion with layout, tables, OCR and RAG-friendly output.
- MinerU — High-quality PDF/document extraction for complex layouts, tables, formulas and OCR.
- PaddleOCR — OCR and document AI toolkit with tables, layout analysis and formulas.
- MarkItDown — Fast general-purpose file-to-Markdown converter from Microsoft.
- AnyDoc — Fast local document conversion to Markdown using a native/WASM-oriented stack.
- PDF Inspector — Fast PDF inspection, classification and extraction/routing.
- Unstructured — Broad document ETL and partitioning toolkit designed for ingestion and RAG.
- OmniParse — Multimodal document parsing for text, images and richer document content.
- Marker — PDF-to-Markdown/JSON parser with strong layout, tables and equation handling.
- OpenDataLoader PDF — PDF-to-structured output with reading order, tables, bounding boxes and benchmark tooling.
- olmOCR — VLM-based OCR and document understanding for difficult and scanned PDFs.
- PyMuPDF4LLM — Lightweight, fast PDF extraction optimized for LLM/RAG workflows.
- GROBID — Scientific-paper parser producing structured TEI/XML, metadata and references.
- Nougat — Academic PDF understanding with strong LaTeX/math and table extraction.
- Apache Tika — General document type detection and text/metadata extraction across many formats.
- Zerox — Vision-model-based document/PDF-to-Markdown extraction.
A curated list of agent harnesses, runtimes, and environments for building, running, and extending AI agents.
- Pi — Minimal and highly extensible agent harness with an agent runtime, tools, sessions, and a coding-agent CLI.
- OpenHarness — Open-source agent harness focused on tools, skills, memory, plugins, permissions, and multi-agent workflows.
- jcode — Coding-agent harness designed around efficient multi-session workflows and extensibility.
- Grok CLI — General-purpose terminal agent from xAI for coding and agentic workflows.
- OpenCode — Open-source terminal coding agent with an extensible architecture and support for multiple model providers.
- Goose — Extensible open-source agent with tools, extensions, and MCP integration.
- OpenHands — Autonomous software-development platform providing agents with coding, terminal, browser, and task-execution capabilities.
- SWE-agent — Agent framework focused on autonomously solving software-engineering tasks and GitHub issues.
- OpenAI Codex CLI — Open-source terminal coding agent with sandboxed execution and tool-driven workflows.
- Gemini CLI — Open-source terminal agent for coding and general agentic tasks with tools and extensibility.
- Cline — IDE-based autonomous coding agent supporting tool execution, MCP, browser interaction, and human-in-the-loop workflows.
- Roo Code — Extensible coding-agent environment supporting custom modes, MCP, instructions, and autonomous execution.
- Aider — Terminal-based AI pair-programming agent with repository-aware context, code editing, and command execution.
- Open Harness — Code-first SDK for building composable AI agents with sessions, tools, middleware, and model providers.
- harness — Unified interface for running different headless coding agents through a common CLI, Python, or TypeScript API.