Open source. Open possibilities.

Astron Agent is an enterprise-grade, open-source agentic workflow platform by iFLYTEK for building production-ready AI agent applications with workflow orchestration, RPA integration, and model management.

Vector is an open-source observability data pipeline written in Rust. It collects, transforms and routes logs and metrics as either an agent or an aggregator, letting teams control where telemetry goes and reduce vendor costs.

17-lesson tutorial teaching harness engineering for AI agents: build the tools, knowledge, permissions, and context management around an LLM agent loop, using Claude Code as the reference design. Runnable Python examples progress from the agent loop to a goal loop.

TiDB is an open-source, cloud-native distributed SQL database with ACID transactions, horizontal scalability, Raft-based high availability, HTAP via TiKV and TiFlash, MySQL 8.0 compatibility, and vector search support.

Excelize is a pure Go library for reading and writing Microsoft Excel spreadsheets in XLAM, XLSM, XLSX, XLTM and XLTX formats. It supports Excel 2007+ files, chart and image insertion, complex spreadsheet components, and a streaming API for large worksheets.

An embedded micro-VM sandbox built on libkrun for safely running AI agents, supporting Linux (KVM) and macOS (Hypervisor.framework).

OpenHands Agent Canvas is a self-hosted developer control center for managing coding agents and automations across local, remote, and cloud backends.

OpenUI is a full-stack, renderer-agnostic Generative UI framework built around OpenUI Lang, a compact streaming-first language that lets AI models generate structured UI with up to 67% fewer tokens than JSON.

UFO³ is Microsoft's open-source framework for multi-device AI agent orchestration, featuring Galaxy for cross-device DAG-based workflows and UFO² for Windows desktop automation, enabling heterogeneous task execution across Windows, Linux, and Android.

An AI production assistant for Ableton Live that helps users understand, arrange, and edit their Live Sets through natural language conversation, with support for multiple AI model connections, audio generation, and stem separation.

Reasonix is an MIT-licensed, DeepSeek-native AI coding agent distributed as a single Go binary. It runs in the terminal, desktop app, browser, or editor via ACP, with config-driven providers, MCP plugins, plan mode, permissions, a workspace sandbox and per-turn checkpoints.

An open-source AI research agent that performs iterative deep research by combining search engines, web scraping, and LLMs. It generates targeted queries, processes results, and recursively explores topics to produce comprehensive markdown reports.

Agent-kernel is an OS-like kernel for multi-agent systems. It provides priority-based scheduling, resource management (including rate limiting and budgets), deadlock detection, and message routing for concurrent AI agents.

FastGPT is a knowledge base platform based on large language models, providing out-of-the-box capabilities such as data processing, RAG retrieval, and visual AI workflow orchestration, supporting rapid construction and deployment of complex Q&A systems.

QwenPaw is a self-hosted personal AI assistant that runs locally or in the cloud, with three-layer memory, sandboxed tool execution, multi-agent support, and connectors for DingTalk, Lark, WeChat, Discord, Telegram and more.

Headroom compresses tool outputs, logs, RAG chunks, files, and conversation history before they reach the LLM, reducing token usage while preserving key information. Runs locally as a Python/TypeScript library, proxy, MCP server, or agent wrapper for Claude Code, Codex, Copilot, Cursor, and more.

OnIt is an AI agent harness that gives a language model tools, a working directory, and memory, then runs the loop until the task is finished. It supports multiple front ends (terminal, browser, A2A, Telegram/Viber) and runs on self-hosted or hosted model servers via MCP tools.

SWE-agent is an autonomous AI agent that uses language models to automatically fix GitHub issues, find cybersecurity vulnerabilities, and solve custom coding challenges. Built by Princeton and Stanford researchers.

Downcity offers AI builders a reusable agent runtime layer: a CLI-hosted local City manages, coordinates and operates many agents, with shared models, tools, memory, tasks, plugins, permissions, usage, billing and Federation services.

LobeHub is an open-source AI agent platform that organizes agents into a 7×24 operation, enabling users to hire, schedule, and collaborate with AI teammates while self-hosting via Docker or Vercel.