#115115th of 144 in Agent frameworks and SDKs
adalflow
PyTorch-style Python library for building and auto-optimizing LLM pipelines and agents
Live demo ↗
(opens in a new tab)Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 4.2k
- License
- MIT
- Last commit
- Nov 2026
- Last release
- Nov 2026
- Language
- Python
Missed the latest GitHub check; these facts may be older.
Overview
AdalFlow is a Python library for building LLM task pipelines (chatbots, RAG, agents) with model-agnostic clients that can be switched via config. Its distinguishing feature is a training loop that treats a workflow as an auto-differentiation graph and optimizes prompts through textual gradients and few-shot bootstrapping. It also ships an Agent and Runner API with sync, async and streaming modes, plus tracing and human-in-the-loop support.
Who it is for: Python developers building LLM pipelines or agents with automated prompt tuning
Strengths
- Textual gradient descent and few-shot bootstrap share one Trainer loop
- Agent Runner offers sync, async and streaming calls with step history and tool events
- Model-agnostic clients; switching providers is a config change
- MIT licensed, installs with pip; tracing has an MLflow integration
Weaknesses
- Optimization needs Parameter, Generator, AdalComponent and Trainer concepts, a steeper learning curve
- Accuracy claims versus other prompt optimizers come from the authors' own research
- Quickstart calculator tool runs eval() on model-generated input, unsafe outside demos
- Library only: no UI, server or deployment tooling
What it needs
- no GPU
- Needs LLM provider API key (OPENAI_API_KEY in the quickstart)
- Models: OpenAI, other providers via model clients
Also in Agent frameworks and SDKs
See all 144| Rank | Project | Score |
|---|---|---|
| 1 | ppt-masterAgent skill that turns documents into natively editable PowerPoint decks | 96 out of 100 |
| 2 | openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing | 92 out of 100 |
| 3 | mastraTypeScript framework for building AI agents, workflows and MCP servers | 92 out of 100 |
| #4 | browser-usePython library that lets an LLM agent control a web browser | 90 out of 100 |
| #5 | copilotkitReact and Angular SDK for adding agent chat and generative UI to apps | 90 out of 100 |