Back to directory

Pocket Flow

A minimalist LLM framework implemented in about 100 lines of Python, providing a single graph abstraction that agents, workflows and RAG pipelines can be built from without vendor-specific wrappers.

Pocket Flow

Pocket Flow is a minimalist LLM framework written in about 100 lines of Python, with no dependencies and no vendor lock-in. It captures the core abstraction shared by larger LLM frameworks in a single graph primitive, and that graph is used to implement common patterns such as (multi-)agents, workflows and retrieval-augmented generation.

It installs as a Python package or can simply be copied as source, and is also available in TypeScript, Java, C++, Go, Rust and PHP. Tutorials cover a chat bot with conversation history, structured-output extraction from resumes, a writing workflow that outlines and styles content, and a research agent, ranging from beginner to more advanced difficulty.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.