Skip to main content
ArkSim connects to your agent through one of three connection types. The examples below show how to wire up each framework.

​
Custom connector

These Python examples load your agent directly as a BaseAgent subclass. No HTTP server needed.

AutoGen

Microsoft AutoGen multi-agent framework.

Claude Agent SDK

Anthropic’s Agent SDK for building agents with Claude.

CrewAI

Role-based multi-agent orchestration.

Google ADK

Google’s Agent Development Kit.

LangChain

LangChain agents and chains.

LangGraph

Stateful, graph-based agent workflows.

LlamaIndex

LlamaIndex agents and query engines.

OpenAI Agents SDK

OpenAI’s SDK for building production agents.

PydanticAI

Type-safe agent framework built on Pydantic.

SmolAgents

Hugging Face’s lightweight agent library.

​
Chat Completions connector

These TypeScript examples expose your agent behind an OpenAI-compatible HTTP endpoint. ArkSim connects using the chat_completions connector.

Mastra

TypeScript agent framework with built-in tool calling.

Vercel AI SDK

Vercel’s SDK for building AI-powered applications.

​
Custom connector with HTTP

This example uses the custom connector with a BaseAgent subclass that forwards requests to a running HTTP server.

Rasa

Open-source conversational AI framework. The custom agent wraps Rasa’s REST webhook channel.
For details on each connection type, see Agent configuration.