All Open Source Apps

Open source AI gateway and Python SDK that gives you one OpenAI-compatible interface to call 100+ LLM providers, with built-in routing, cost tracking, guardrails, and virtual keys.

Build, test, and deploy AI agents and RAG workflows visually with native API and MCP server export.

The batteries-included Python agent harness — planning, sub-agents, filesystem, shell, memory, and skills bundled in, built on LangGraph.

Local-first AI memory with verbatim storage, pluggable backends, and 96.6% retrieval recall on LongMemEval — no API key required.

A Microsoft-built training framework that optimizes AI agents with reinforcement learning, automatic prompt optimization, or supervised fine-tuning — with near-zero code changes to your existing agent, in any framework.

Fast, stable Redis GUI with cluster, SSH, and massive key support

An all-in-one open-source training suite for finetuning diffusion models—FLUX, SDXL, Wan video, and audio—on consumer GPU hardware via CLI or web UI.

A fully local web research assistant that iteratively searches, summarizes, and refines markdown reports using any Ollama or LMStudio model—no cloud or API keys required.

Build autonomous and human-in-the-loop multi-agent AI systems with a layered, event-driven Python and .NET framework pioneered at Microsoft Research.

An Apple-Silicon-native LoRA fine-tuning tool for Gemma on text, image, and audio data — with a wizard CLI, live browser-based training visualizer, and streaming from GCS/BigQuery for datasets too large for local disk.

Baidu's open-source OCR model for "one-shot, long-horizon parsing" — published with an arXiv paper, a Hugging Face model release, and vLLM inference support, aimed at document understanding beyond typical page-by-page OCR limits.

An open-source build system that turns written specs and architecture into working code — an LLM generates code under tight constraints, task by task with narrow context windows, instead of attempting an entire codebase at once.

Give an AI agent a real LLM training setup and let it experiment autonomously overnight — you wake up to a log of experiments and (hopefully) a better model.