HomeToolsMCPHow It WorksStoriesPhilosophyArchitectureStar on GitHub
All Tools
P
AgentsFreeOpen Source

POCKETFLOW

The 100-line LLM framework — minimal, composable, zero lock-in

MIT

ABOUT

Existing LLM frameworks like LangChain and CrewAI are bloated with tens of thousands of lines, heavy dependencies, and vendor-specific wrappers that create lock-in and maintenance burdens. PocketFlow provides just the essential graph abstraction (Node, Flow, SharedStore) in roughly 100 lines of zero-dependency Python, eliminating framework bloat while still supporting advanced patterns like multi-agent orchestration, RAG pipelines, and conditional workflows.

INSTALL
pip install pocketflow

INTEGRATION GUIDE

1. Build autonomous AI agents that make decisions, use tools, and loop until tasks are complete 2. Implement RAG pipelines that integrate document retrieval with LLM generation for grounded Q&A 3. Orchestrate multiple specialized AI agents that collaborate on complex multi-step tasks 4. Create conditional workflow pipelines that branch and route based on LLM outputs 5. Rapidly prototype LLM applications using the minimal framework without vendor lock-in

TAGS

pythonllm-frameworkminimalagentsragworkflowagentic-aigraphzero-dependencycomposable