tools: LangGraph: AI agent with state machine, and similar

The state machine + agent pattern is powerful, and LangGraph has popularized it, but it's not unique to them.

Similar Approaches

Direct competitors with explicit graph/state concepts:

• Autogen (Microsoft) - Uses a different model (multi-agent conversation) but has added StateFlow for explicit state machine control
• CrewAI - More task/role-based but recently added "Flows" which provide state machine semantics
• Rivet (Ironclad) - Visual node-based graph editor for AI workflows, compiles to JSON that can run from any language
• Llama Index Workflows - Their newer workflow system has explicit state and conditional branching

Workflow engines adding AI capabilities:

• Temporal - You can absolutely build agent loops with Temporal activities; it gives you durability and fault tolerance that LangGraph lacks out of the box
• Inngest - Similar to Temporal but lighter weight, good for AI agent patterns
• Prefect / Dagster - Data pipeline tools that can orchestrate agent-style flows

Custom - you could implement the same patterns. The core insight is just: state + transitions + conditional routing based on LLM outputs. Step Functions with a "loop until done" pattern and dynamic choice states.

LangGraph Language Support

Native SDKs:

• Python (primary, most features)
• JavaScript/TypeScript (solid, slightly behind Python in features)

LangGraph Platform (Cloud/Self-hosted):

• Exposes a REST API, so you can call it from any language
• But you still need to define your graphs in Python or JS

So if you wanted to orchestrate from, say, a .NET service, you'd either define the graph in Python and call the LangGraph API

LangGraph

• Website: https://www.langchain.com/langgraph
• GitHub: https://github.com/langchain-ai/langgraph
• Docs: https://docs.langchain.com/langgraph-platform
• Languages: Python, JavaScript/TypeScript (native SDKs); REST API for any language

Similar Tools / Competitors

AutoGen (Microsoft)

• Website: https://www.microsoft.com/en-us/research/project/autogen/
• GitHub: https://github.com/microsoft/autogen
• Docs: https://microsoft.github.io/autogen/stable/
• Languages: Python, C#/.NET

CrewAI

• Website: https://www.crewai.com/
• GitHub: https://github.com/crewAIInc/crewAI
• Docs: https://docs.crewai.com
• Languages: Python only

Rivet (Ironclad)

• Website: https://rivet.ironcladapp.com/
• GitHub: https://github.com/Ironclad/rivet
• Docs: https://rivet.ironcladapp.com/docs
• Languages: Visual editor exports to JSON; TypeScript library for runtime (@ironclad/rivet-core, @ironclad/rivet-node)

LlamaIndex Workflows

• Website: https://www.llamaindex.ai/workflows
• GitHub: Part of LlamaIndex core
• Docs: https://docs.llamaindex.ai/en/stable/module_guides/workflow/
• Languages: Python, TypeScript

Workflow Engines (general-purpose, AI-capable)

Temporal

• Website: https://temporal.io/
• AI Landing Page: https://temporal.io/solutions/ai
• GitHub: https://github.com/temporalio/temporal
• Languages: Go, Python, Java, TypeScript, .NET, Ruby
• Note: OpenAI's Codex agent and Replit's Agent 3 are built on Temporal Temporal

Inngest

• Website: https://www.inngest.com/
• AI Landing Page: https://www.inngest.com/ai
• GitHub: https://github.com/inngest/inngest
• Languages: TypeScript/JavaScript, Python, Go

Who's trying similar things:

Tool	Approach	Gap vs LangGraph
CrewAI Flows	Event-driven with @listen, @router decorators	Less explicit graph structure, more implicit event wiring
LlamaIndex Workflows	@step + events, state via Context	Similar feel, but events rather than edges — less visual
Semantic Kernel Process Framework	Stateful processes, Dapr/Orleans for distribution	More enterprise/Microsoft stack oriented
Rivet	Visual graph, but agents are nodes not routers	Agent doesn't decide routing — you wire it statically
Temporal	You can build it, but it's DIY	No native "LLM picks next step" primitive