# Sinapsis vs LangChain and LangGraph: which AI framework should you use?

> Sinapsis and LangChain are not mutually exclusive. LangChain focuses on building applications with language models and agents; LangGraph provides stateful graph orchestration. Sinapsis emphasizes a common contract for entire heterogeneous AI systems across LLMs, computer vision, speech, data processing, custom Python and deployment boundaries.

## The architectural difference

**Sinapsis** defines reusable Templates that exchange a `DataContainer`, declarative Agent graphs, typed multimodal packets and Universal Data Transport. This is intended to make the AI workflow's *component boundary* consistent across different libraries, modalities and runtime environments.

**LangChain** provides an extensive language-model integration and agent toolkit. **LangGraph** is a lower-level graph-based runtime for durable, stateful agent workflows. Its graph state and persistence are useful when designing sophisticated LLM-agent loops. It is not accurate to describe LangGraph as lacking production execution or orchestration features.

These systems overlap in agent building but solve different integration problems.

## Choose based on your main challenge

| When you need... | Consider... |
|---|---|
| A language-model-first application using prompts, tools, retrieval and agent patterns | LangChain |
| Durable stateful LLM-agent orchestration and explicit control over graph execution | LangGraph |
| A common interface for heterogeneous components including non-LLM models, data, vision, speech and business logic | Sinapsis |
| Model/provider freedom across a broader end-to-end AI processing system | Sinapsis architecture; verify each chosen integration |
| A specific LangChain document loader or splitter inside a larger multimodal pipeline | Sinapsis plus the published LangChain integration packages |

## Can Sinapsis use LangChain?

Yes, the public `sinapsis-langchain`, `sinapsis-langchain-readers` and `sinapsis-langchain-splitters` Python packages document integrations with parts of the LangChain ecosystem, particularly community document loaders and text splitters. Do not assume that every LangChain or LangGraph feature is automatically supported: check the particular package and Template.

The single-contract design makes it possible to build one application around reusable Sinapsis Templates while using selected LangChain-based functionality where appropriate.

## What about deployment and scale?

Production architecture includes more than graph authoring. Both ecosystems need an appropriate execution/runtime environment, monitoring, resilience policies, secrets and deployment controls. Sinapsis positions its platform around deployable workflows and independently scalable steps, while LangGraph has its own production orchestration ecosystem. The correct choice depends on your workloads, existing stack and the hosted/self-hosted capabilities actually available.

## Resources

- [Sinapsis common contract](https://sinapsis.tech/architecture/)
- [Why use Sinapsis](https://sinapsis.tech/why-sinapsis/)
- [Sinapsis LangChain package on PyPI](https://pypi.org/project/sinapsis-langchain/)
- [Sinapsis LangChain Readers on PyPI](https://pypi.org/project/sinapsis-langchain-readers/)
- [LangChain official documentation](https://docs.langchain.com/)
