# Introduction

> What 10xGraph is, what it generates for you, and where to start. An open-source Python framework for production multi-agent AI.

Source: https://10xgraph.com/docs
Last updated: 2026-10-06

10xGraph is an open-source Python framework for production multi-agent AI. You write the agent, and 10xGraph generates the production server around it: authentication, scoped access control, owner-only threads, rate limits, and Docker and Kubernetes deployment. Memory has two tiers, with hot data in a Redis cache and cold data in PostgreSQL.

## What you write

A graph of agents and tools in plain Python. Use the prebuilt `ReactAgent` for the common tool-calling loop, or build your own `StateGraph` when you need routing, parallel branches or human approval steps. Tools are ordinary Python functions, such as `lookup_order(order_id: str)` or `refund_order(order_id: str, amount: float)`. Models come from OpenAI, Google Gemini or Anthropic, and you can change the model without touching the graph.

When you are ready to serve it, you point `10xgraph.json` at your compiled graph. You do not write routes, auth middleware or deployment files by hand.

## What 10xGraph generates

| Piece | What you get |
|---|---|
| API server | REST, SSE streaming, WebSocket and realtime audio endpoints on port 8000 |
| Security | JWT or custom auth, owner-only threads, role scopes on every endpoint, rate limits |
| Reliability | Replay-safe tools, versioned state writes, node and tool timeouts |
| Deployment | Dockerfile, docker-compose.yml and a Kubernetes Deployment and Service |

> **Formerly Agentflow**
> 10xGraph was published as Agentflow until 2026. The Python package is now imported as `tenxgraph`, as in `from tenxgraph import StateGraph`. The old `agentflow` import stays available as a deprecated alias until 2.0.

## Where to go next

- [Quickstart](https://10xgraph.com/docs/get-started/first-agent): Build an agent with a tool, run it, then serve it over HTTP.
- [Project structure](https://10xgraph.com/docs/get-started/project-structure): Every file the production template generates and what it does.
- [Replay-safe tools](https://10xgraph.com/docs/concepts/replay-safe-tools): How a resumed run skips tools that already finished.
- [Memory: hot and cold](https://10xgraph.com/docs/concepts/memory): Redis for the hot path, PostgreSQL for durable history.
- [Add JWT authentication](https://10xgraph.com/docs/how-to/api-cli/add-auth): Turn on auth and owner-only threads in 10xgraph.json.
- [CLI reference](https://10xgraph.com/docs/reference/api-cli/commands): Every command and flag of the 10xgraph CLI.

## Build something real

Each use case page gives a reference architecture, the tools it needs and the guardrails to add before it ships.

- [Customer support agent](https://10xgraph.com/docs/use-cases/customer-support-agent): Intent routing, order lookup, refund tools and human handoff on escalation.
- [Data extraction agent](https://10xgraph.com/docs/use-cases/data-extraction-agent): Pull structured fields from unstructured text, validate them and retry on errors.
- [Coding agent](https://10xgraph.com/docs/use-cases/coding-agent): Plan first, then read files, edit with diff approval and run tests.
- [Research agent](https://10xgraph.com/docs/use-cases/research-agent): Split a question into searches, fetch sources and enforce citations.
- [RAG agent](https://10xgraph.com/docs/use-cases/rag-agent): Agentic retrieval with hybrid search, reranking and forced citations.

## Integrate with your stack

10xGraph fits into a service or frontend you already run. It does not ask you to replace it.

- [FastAPI](https://10xgraph.com/docs/integrations/agentflow-with-fastapi): Embed an agent in an existing FastAPI app, or run the API server beside it.
- [Next.js](https://10xgraph.com/docs/integrations/agentflow-with-nextjs): Stream tokens from a Python agent into a Next.js frontend with auth and typed responses.
- [CopilotKit](https://10xgraph.com/docs/integrations/agentflow-with-copilotkit): Serve the agent over the AG-UI endpoint and connect a CopilotKit chat.
- [Postgres](https://10xgraph.com/docs/integrations/agentflow-with-postgres): Durable agent threads with PgCheckpointer on Postgres, with Redis as the hot cache.
- [Model providers](https://10xgraph.com/docs/providers): OpenAI, Google Gemini and Anthropic, plus any OpenAI-compatible endpoint.

## Ship to production

Start with the build guide, which takes a support agent from code to a secured API with tests and Docker files. Then use the how-tos for each concern.

- [Production ReAct agent](https://10xgraph.com/build/production-react-agent): A support agent that looks up and refunds orders, served with JWT auth, rate limits and Postgres memory.
- [Deployment](https://10xgraph.com/docs/how-to/production/deployment): Containers, runtime settings, shared persistence and release checks.
- [Docker and Kubernetes](https://10xgraph.com/docs/how-to/api-cli/generate-docker-files): Generate a Dockerfile, docker-compose.yml and Kubernetes manifest with 10xgraph build.
- [Kubernetes](https://10xgraph.com/docs/how-to/production/kubernetes): Set grace periods, probes and scaling so rolling deploys do not cut off a run.
- [Auth and authorization](https://10xgraph.com/docs/how-to/production/auth-and-authorization): Secure the API with JWT or a custom backend and scope access on every endpoint.
- [Rate limiting](https://10xgraph.com/docs/how-to/api-cli/configure-rate-limiting): Turn on the built-in sliding-window limiter, in memory or on Redis.
- [Checkpointing](https://10xgraph.com/docs/how-to/production/checkpointing): Choose and configure a checkpointer so threads survive restarts.

## How the docs are organized

Each section answers a different question. Use the list to jump to the one you need.

- [Get started](/docs/get-started): install 10xGraph, build a first agent and see what the production template generates.
- [Beginner path](/docs/beginner): a guided route from the mental model to a served agent and a TypeScript client.
- [Concepts](/docs/concepts): how graphs, state, tools, memory and the production runtime work.
- [Prebuilt](/docs/prebuild): ready-made agents and tools to use as they are or extend.
- [How-to guides](/docs/how-to): task recipes for the Python library, the CLI, the TypeScript client and production.
- [Testing and QA](/docs/qa): unit tests with mocked models, evaluation sets and simulated users.
- [Tutorials](/docs/tutorials): end-to-end builds based on the examples in the repository.
- [Reference](/docs/reference): exact details for the Python library, REST API, CLI and client.
- [Troubleshooting](/docs/troubleshooting): symptoms, causes and fixes, plus the error code reference.
- [Learn more](/docs/learn-more): use cases, integrations, providers and framework comparisons.
- [Glossary](/docs/glossary): plain definitions of AI agent terms, from ReAct to durable execution.
- [Project](/docs/project): roadmap, security policy, upgrade guides, support and contributing.

## Pick a path

**New to agents.** Read the [beginner mental model](/docs/beginner/mental-model), then build [your first agent](/docs/beginner/your-first-agent) and [add a tool](/docs/beginner/add-a-tool).

**Already using LangGraph.** Start with [10xGraph vs LangGraph](/docs/compare/agentflow-vs-langgraph) for an honest side-by-side, including where LangGraph is stronger, such as its visual tooling. Then read [Concepts](/docs/concepts) to see how `StateGraph` maps across, and the [Quickstart](/docs/get-started/first-agent) to run something. 10xGraph is pre-1.0, so pin versions and read the changelog.

**Going to production.** Begin with [production how-to guides](/docs/how-to/production) for [auth and authorization](/docs/how-to/production/auth-and-authorization), [checkpointing](/docs/how-to/production/checkpointing) and [deployment](/docs/how-to/production/deployment). Then read [Replay-safe tools](/docs/concepts/replay-safe-tools) before you give an agent tools that move money or send email, and keep [Troubleshooting](/docs/troubleshooting) close when you deploy.

## Frequently asked questions

### Is 10xGraph production-ready?

10xGraph is pre-1.0, so pin versions and read the changelog before each upgrade. The production layer (API server, JWT or custom auth, owner-only threads, rate limits, replay-safe tools, Docker and Kubernetes files) ships in the same install, and 10xScale runs it for its own AI products.

### Do I need LangChain to use 10xGraph?

No. 10xGraph has no LangChain dependency. You build graphs with its own StateGraph and Agent classes and call model providers through its own interface.

### Which models does 10xGraph support?

OpenAI, Google Gemini (including Vertex AI) and Anthropic (direct API, Vertex AI and Amazon Bedrock), plus any OpenAI-compatible endpoint such as Ollama or vLLM. Swapping the model string does not change your graph or tools.

### Can I self-host 10xGraph?

Yes. 10xGraph is MIT licensed and self-hosted, with no hosted platform required. The build command writes a Dockerfile, docker-compose.yml and a Kubernetes manifest so you can run it on your own infrastructure.