# Agentmatic > A drop-in replacement for LangGraph. Same API. Rust engine. 10x faster. Agentmatic is an open-source AI agent framework — drop-in compatible with LangGraph, built on a Rust engine. Native SDKs for Python, TypeScript, Rust, Go and Java. Built-in circuit breakers, retry, dead letter queues, distributed clusters and checkpointing. MIT-licensed. ## Key URLs - [Home](https://agentmatic.neullabs.com/): Marketing home with hero, features, benchmarks, FAQ. - [Migrate from LangGraph](https://agentmatic.neullabs.com/migrate): Step-by-step migration guide. - [Benchmarks](https://agentmatic.neullabs.com/benchmarks): Performance comparison vs LangGraph. - [Architecture](https://agentmatic.neullabs.com/architecture): Rust engine + multi-SDK design. - [Resilience](https://agentmatic.neullabs.com/resilience): Circuit breakers, retry, dead-letter queues. - [Distributed](https://agentmatic.neullabs.com/distributed): Self-hosted coordinator/worker clusters. - [Checkpointing](https://agentmatic.neullabs.com/checkpointing): Time-travel debugging + HITL. - [MCP](https://agentmatic.neullabs.com/mcp): Model Context Protocol integration. - [Pricing](https://agentmatic.neullabs.com/pricing): MIT-licensed, free forever; optional paid support. - [FAQ](https://agentmatic.neullabs.com/faq): Common questions. - [GitHub](https://github.com/neul-labs/agentmatic): Source code, issues, releases. - [Docs](https://docs.neullabs.com/agentmatic): Technical documentation. ## Articles - [Build a code review agent in 2 hours with MCP and ReAct](https://agentmatic.neullabs.com/blog/build-code-review-agent-with-mcp-react): End-to-end walkthrough: build an AI code reviewer using Agentmatic's MCP client to wrap the filesystem + git MCP servers, a ReAct loop, and a custom GitHub-comment tool. Production-ready. - [CrewAI vs AutoGen vs Agentmatic: multi-agent paradigms head-to-head](https://agentmatic.neullabs.com/blog/crewai-vs-autogen-vs-agentmatic): Three different bets on how multi-agent systems should be built — role-based crews, conversation orchestration, or explicit state graphs. The honest tradeoffs. - [Agent checkpointing, time travel, and human-in-the-loop done right](https://agentmatic.neullabs.com/blog/agent-checkpointing-time-travel-hitl): How checkpointing actually works in Agentmatic, why every superstep is a save point, how time-travel debugging changes development, and the HITL patterns that scale to production. - [LangChain tools in a Rust-engined agent: the zero-friction bridge](https://agentmatic.neullabs.com/blog/langchain-tools-in-rust-zero-friction-bridge): How from_langchain_tools() actually works, why bringing your existing LangChain toolbox costs zero engineering time, and the cases where you should rewrite a tool natively anyway. - [Distributed AI agents: self-hosted at scale without a SaaS dependency](https://agentmatic.neullabs.com/blog/distributed-ai-agents-self-hosted-at-scale): How to run an AI agent runtime across a cluster, in your VPC, without paying a platform tier. Coordinator/worker, P2P mesh, checkpoint hand-off, observability — all from open-source primitives. - [ReAct vs Supervisor vs RAG: choosing the right agent pattern](https://agentmatic.neullabs.com/blog/react-vs-supervisor-vs-rag-patterns): The three most common AI agent patterns — when each fits, when each fails, and how to compose them. Code in every Agentmatic SDK. - [Production AI agents need circuit breakers, retry, and dead-letter queues](https://agentmatic.neullabs.com/blog/production-ai-agents-circuit-breakers-retry-dlq): The three resilience primitives every production agent eventually needs — and the mistakes engineers make when they bolt them on themselves. Plus what 'built-in' looks like in Agentmatic. - [Model Context Protocol (MCP) for AI agents: a complete guide](https://agentmatic.neullabs.com/blog/mcp-model-context-protocol-guide): What MCP is, why it became the open standard for AI agent tools, how to consume MCP servers from your agent, how to ship your own MCP server, and the gotchas nobody mentions. - [Best AI agent frameworks in 2026: a working developer's comparison](https://agentmatic.neullabs.com/blog/best-ai-agent-frameworks-2026): An opinionated look at LangGraph, Agentmatic, CrewAI, AutoGen, Vercel AI SDK, OpenAI Agents SDK and friends. Not 'which is best' — 'when to pick which'. - [LangGraph vs Agentmatic: a careful performance comparison](https://agentmatic.neullabs.com/blog/langgraph-vs-agentmatic-performance): An honest, reproducible LangGraph vs Agentmatic benchmark across ReAct, Supervisor, RAG, and Map fan-out workloads. With methodology, hardware specs, raw numbers, and the cases where the gap is small. - [Migrate from LangGraph to Agentmatic in 5 minutes](https://agentmatic.neullabs.com/blog/migrate-from-langgraph-in-5-minutes): A real, no-handwaving step-by-step LangGraph → Agentmatic migration: install, change one import, validate with your existing tests, and optionally adopt resilience primitives. Plus the actual gotchas. - [The fastest LangGraph alternative is LangGraph — recompiled in Rust](https://agentmatic.neullabs.com/blog/langgraph-alternative-rust-engine): We rebuilt the LangGraph runtime in Rust, kept the Python API identical, added circuit breakers and distributed clusters in the open-source core, and shipped Python, TypeScript, Rust, Go, and Java SDKs. Here's why. ## Comparisons - [Agentmatic vs CrewAI — graph state machines vs role-based crews](https://agentmatic.neullabs.com/compare/crewai): CrewAI is role-based delegation (agents play roles, complete tasks). Agentmatic is graph-based state machines (deterministic, parallel, checkpointable). Different paradigms — and they can be complementary. - [Agentmatic vs LangGraph — drop-in replacement, 10× faster](https://agentmatic.neullabs.com/compare/langgraph): Agentmatic is API-compatible with LangGraph and runs on a Rust engine. Same StateGraph, same checkpointers, same patterns — 10–15× faster traversal, plus circuit breakers and distributed clusters in the open-source core. - [Agentmatic vs AutoGen — conversation-driven vs state-graph orchestration](https://agentmatic.neullabs.com/compare/autogen): Microsoft AutoGen orchestrates conversations between agents. Agentmatic orchestrates state graphs. Different paradigms, different sweet spots — both viable for multi-agent systems. - [Agentmatic vs OpenAI Agents SDK — open framework vs vendor-led toolkit](https://agentmatic.neullabs.com/compare/openai-agents-sdk): OpenAI's Agents SDK is a clean toolkit for building agents against OpenAI's API. Agentmatic is an open framework against any LLM, with explicit graphs, multi-language SDKs, and production resilience primitives. - [Agentmatic vs Vercel AI SDK — orchestration framework vs LLM toolkit](https://agentmatic.neullabs.com/compare/vercel-ai-sdk): Vercel AI SDK is an excellent LLM client + streaming + tool-calling toolkit. Agentmatic is an agent orchestration framework. They're complementary, not competitive — and we ship an adapter. ## Recipes - [Run an agent across a distributed cluster](https://agentmatic.neullabs.com/recipes/distributed-cluster): Coordinator + worker (or P2P mesh) over gRPC, with least-loaded routing and checkpoint hand-off. Ship the same agent code to a cluster without rewriting it. - [Wrap an MCP server as Agentmatic tools](https://agentmatic.neullabs.com/recipes/mcp-tool): Connect to any MCP (Model Context Protocol) server and expose its tools to your agent. Three lines for stdio servers, four for SSE. - [Human-in-the-loop approval before risky tool calls](https://agentmatic.neullabs.com/recipes/hitl-approval): Pause the graph before a sensitive tool fires, surface the proposed call to a human, then resume with their decision. Used for refunds, prod DB writes, transfers. - [The RAG pattern: retrieval-augmented generation](https://agentmatic.neullabs.com/recipes/rag-pattern): A retrieval-augmented generation agent: embed the question, look up relevant chunks, synthesize a grounded answer with citations. The right default for Q&A over your docs. - [The ReAct pattern: reason-then-act with tool calling](https://agentmatic.neullabs.com/recipes/react-pattern): A reusable ReAct (reason + act) loop for tool-calling agents. Includes the prebuilt create_react_agent helper plus a hand-rolled StateGraph version in 5 languages. - [The Supervisor pattern: routing to specialist sub-agents](https://agentmatic.neullabs.com/recipes/supervisor-pattern): A supervisor agent that reads the user message and routes to specialist sub-agents (docs, billing, escalation). The default pattern for multi-domain customer-facing agents. ## Examples - [Code reviewer — MCP filesystem + git + ReAct](https://agentmatic.neullabs.com/examples/code-reviewer): An AI code reviewer that connects to local MCP servers (filesystem + git), reads diffs, and posts structured review comments. Ships in Python, TypeScript and Rust. - [Support bot — supervisor + RAG + ReAct + HITL](https://agentmatic.neullabs.com/examples/support-bot): A production-grade customer support agent that routes between specialist sub-agents, retrieves docs, escalates to humans on refunds, and runs in 5 SDKs. - [Research agent — sandboxed Rhai scripts as LLM tools](https://agentmatic.neullabs.com/examples/research-agent): An LLM-driven research agent that writes and executes Rhai scripts in a deterministic sandbox. Lets the model use code without giving it the keys to the kingdom. - [Doc pipeline — custom Pregel graph with map fan-out](https://agentmatic.neullabs.com/examples/doc-pipeline): A high-throughput document processing pipeline that fans out to N parallel workers, with no LLM. Shows raw Pregel API in Rust — extract, dedup, embed, persist.