FareedKhan-dev/all-agentic-architectures

What is it?

Key features

• 01Implements 35 agentic patterns behind a uniform run(task) interface.
• 02Supports nine LLM providers, switchable via environment variable.
• 03Each architecture has a runnable, executed Jupyter notebook.
• 04Includes a 17-task benchmark comparing architectures on the same tasks.
• 05Built on LangGraph state machines.
• 06Includes a BrowserAgent that drives a real Playwright browser with a safety gate.

Use cases

• Learning how reflection, search, RAG and multi-agent patterns differ in practice.
• Prototyping an agent workflow and comparing candidate architectures before committing.
• Switching LLM providers without rewriting agent code.
• Adding a new agent pattern to a shared, tested interface.

Quick start

1. 1

   Install the library with extras

   pip install "agentic-architectures[nebius,faiss,tavily]"
2. 2

   Run a Reflection architecture on a task using get_llm

Things to know

• !Default examples need a paid LLM provider API key such as Nebius, OpenAI or Anthropic.
• !Benchmark results come from one model run and show only 78% correct overall.
• !Full test suite and provider setup require a virtualenv and a .env file with keys.

Topics

FAQ

What is all-agentic-architectures?

A Python library and runnable Jupyter textbook implementing 35 agentic AI architectures, from Reflexion and ReAct to MemGPT and BrowserAgent, on LangGraph. This project packages a broad set of agentic AI patterns from the research literature, such as Reflexion, Tree of Thoughts, LATS, GraphRAG, MemGPT, Voyager and STORM, as runnable Architecture classes. Each class exposes the same run(task) interface and returns the same result shape, so swapping one pattern for another does not require changes to calling code.

Is all-agentic-architectures free to use? What license does it have?

all-agentic-architectures is open source under the MIT license. Check the license terms before using it in commercial products.

Is all-agentic-architectures still maintained?

Updated within 6 months. The last commit to all-agentic-architectures was on Jun 22, 2026.

What can I use all-agentic-architectures for?

Learning how reflection, search, RAG and multi-agent patterns differ in practice. Prototyping an agent workflow and comparing candidate architectures before committing. Switching LLM providers without rewriting agent code. Adding a new agent pattern to a shared, tested interface.

How do I install all-agentic-architectures?

Install the library with extras: pip install "agentic-architectures[nebius,faiss,tavily]" Run a Reflection architecture on a task using get_llm

Where can I get all-agentic-architectures?

The source code and installation instructions are on GitHub: https://github.com/FareedKhan-dev/all-agentic-architectures