activity-doc - HackMD

# ai-lab-activity-20250502 let's experiment with a few of the popular low-code / no-code ways of building agents and agentic systems. ## GPT-assisted list of options Below is a list of seven no-code AI agent-building platforms—Crew AI, Langflow, Botpress, n8n, Bubble, Lindy, and Stack AI—highlighting each tool’s visual interface, integration ecosystem, template libraries, and suitability for rapid, hands-on experimentation ([crewai.com][1], [Langflow][2], [Botpress][3], [n8n][4], [Bubble][5], [Lindy — Meet Your AI Assistant][6], [Stack AI][7]). ### Voiceflow Voiceflow gives teams a drag-and-drop canvas and collaborative Workflow Builder for designing voice, chat, and multimodal agents side-by-side in real time. Its GPT-powered **AI Builder** layers prompt-chaining, auto-generated dialogue, and built-in conversation memory, so sophisticated flows come together in minutes without code. When deeper control is needed, developers can “bring your own LLM” through a dedicated Custom LLM step or call any external service with the API step, blending low-code speed with full programmability. More than 300 one-click integrations and a template marketplace accelerate deployment to web widgets, telephony, or popular messaging channels. Finally, a native analytics dashboard tracks token usage, interaction counts, latency, and other KPIs, giving product teams the data they need to iterate and improve every release. ### Crew AI Crew AI provides a drag-and-drop “UI Studio” for assembling multi-agent workflows, known as “crews,” using prebuilt templates and a testing sandbox for iterative improvement ([crewai.com][1], [Analytics Vidhya][8]). It’s optimized for orchestrating multiple agents in parallel or sequence, making it ideal for experiments in coordination patterns. ### Langflow Langflow is a visual, low-code builder focused on Retrieval-Augmented Generation (RAG) and multi-agent workflows. Users connect nodes representing models, APIs, or data sources to craft end-to-end pipelines without writing code ([Langflow][2], [DataStax][9]). The recent Langflow Desktop release adds an offline GUI for designing and testing agents locally ([Langflow][10]). ### Botpress Botpress offers a complete no-code platform for conversational agents, featuring a drag-and-drop flow editor, built-in NLU, analytics dashboards, and multi-channel deployment (e.g., WhatsApp, Slack) ([Botpress][3], [Botpress][11]). Its plugin architecture also allows gradual extension with code if needed. ### n8n n8n is an open-source workflow automation tool with an “AI Agent” integration that lets you call LLMs alongside 400+ app connectors in visual pipelines ([n8n][4], [n8n Blog][12]). This makes it easy to prototype agents that ingest data, apply AI logic, and output results to nearly any service. ### Bubble Bubble’s no-code web app builder includes AI agent templates and plugins, enabling you to design custom chatbots and data-driven assistants via visual workflows and natural language prompts ([Bubble][5], [Bubble][13]). It’s well suited for users who want full control over front-end UI and AI logic in one environment. ### Lindy Lindy provides a streamlined studio for creating self-learning AI agents (“assistants”) that automate workflows across apps like Gmail, HubSpot, and more within minutes ([Lindy — Meet Your AI Assistant][6], [Lindy — Meet Your AI Assistant][14]). Agents can communicate with each other and continuously improve via built-in testing and training tools, backed by SOC 2 and HIPAA compliance ([Lindy — Meet Your AI Assistant][15]). ### Stack AI Stack AI is an enterprise-grade, drag-and-drop platform for building AI applications and agents. It offers templates for chatbots, simulations, and even Canvas LMS integration (as piloted at MIT Sloan), all without coding ([Stack AI][7], [MIT Sloan Tech & Learning][16]). Backed by low-code hooks and YC backing, it balances ease of use with enterprise scalability ([Y Combinator][17]). ## Emerging Trends & Considerations * **Hybrid Low-Code/No-Code**: Platforms increasingly offer code hooks alongside visual builders for advanced customization ([Lindy — Meet Your AI Assistant][14]). * **Multi-Agent Orchestration**: Native support for agent collaboration, handoff patterns, and monitoring. * **Industry-Specific Templates**: Pretrained blueprints for customer support, sales outreach, and research assistants. * **Responsible AI & Compliance**: Built-in audit logs, data encryption, and adherence to GDPR, SOC 2, and HIPAA standards. ## "High Code" **Searching for AI agent frameworks** It looks like I'm trying to find AI agent frameworks that are high-code or developer-focused. I'll search for terms like "LangChain," "LangGraph," "OpenAI Agents SDK," "Autogen," and others. These seem to relate to frameworks or libraries for building AI agents. I’ll start off by checking for general agent frameworks in 2025 and proceed with specific searches to make sure I track down the right libraries and resources, especially looking for official pages. Let’s get started! **Refining AI agent framework search** I’m narrowing down my search to high-code frameworks, especially things like LangChain, LangGraph, and OpenAI Agents SDK. I’ll also check for Microsoft’s Semantic Kernel, Haystack (for question answering), and possibly other frameworks like AutoGen and LlamaIndex. I’ll make sure to focus on agent frameworks that are developer-oriented. I'll go ahead and start searching for these terms to find a solid list of options, including a few wildcards like Ray Serve and AgentFlow. Let’s dig into it! **Running AI agent framework searches** I want to search different AI agent frameworks. These include "high-code AI agent frameworks," "LangGraph," "OpenAI Agents SDK," "Microsoft Semantic Kernel," "Autogen," "Haystack," and "Jina". Each search will be done with an emphasis on getting results that are up-to-date. The idea is to find a range of relevant information, so I’ll run about eight searches to cover all my bases. Let's go ahead and start! **Searching for LangGraph details** It seems like I’m weighing the choice between looking at getzep.com and Reddit for LangGraph information, but I’m leaning toward official documentation. I’ll need to verify LangGraph’s details by searching "LangGraph," "langgraph," or checking the GitHub repository for the most reliable sources. That should provide some clear, credible information. I’ll focus on official documentation over blog posts for accuracy. Let’s dig into that search and see what comes up! **Searching for AI agent frameworks** I’ll search for a list of high-code AI agent frameworks, like LangChain, LangGraph, OpenAI Agents SDK, and Microsoft’s Semantic Kernel. I’ll also look into AutoGen, Haystack, and Jina to see if they fit in the agent framework category. I found some relevant pages for LangChain, LangGraph, and OpenAI Agents SDK, including product pages and discussions, so I’ll collect the info from those. Let’s run these searches and gather everything needed for a solid answer. **Identifying frameworks for agents** I’m considering including high-code frameworks like LangChain, LangGraph, and OpenAI Agents SDK, which seem like strong candidates. However, I’m unsure about including LlamaIndex since it’s more focused on indexing rather than agent orchestration. Still, it's a framework for building retrieval and chaining, so it could be argued. I’ll focus on clear agent orchestrators—LangChain, LangGraph, and the OpenAI Agents SDK—and consider whether Microsoft’s Semantic Kernel fits in well. Let’s keep it to high-code industry standards. **Identifying high-code frameworks** I've narrowed down the high-code frameworks to include: * **LangChain**: Referencing its official page. * **LangGraph**: Also from the GitHub. * **OpenAI Agents SDK**: Using OpenAI’s SDK page. * **Microsoft Semantic Kernel**: Cited from Microsoft Learn and GitHub. There's also **LlamaIndex**, but it focuses more on retrieval than agent orchestration. While it can work like an agent, I'll leave it out for now. I'll also consider **Haystack**, which has added agent features in 2024. **Listing high-code agent frameworks** Here's a list of frameworks that require high-code development: * **LangChain** * **LangGraph** * **OpenAI Agents SDK** * **Microsoft Semantic Kernel** * **Haystack** (with agents) * **Rasa** I’m also considering **Jina AI**, which has an Executor for LLM, though it’s more focused on retrieval and not strictly agent orchestration. **Agentic frameworks** by Google DeepMind might also be relevant, and **AgentFlow** could be an option. But for now, the core frameworks will focus on the ones listed above. **Considering agent frameworks** It seems like the user wants industry-standard, high-code agent frameworks. So, I should focus on LangChain, LangGraph, OpenAI Agents SDK, Microsoft Semantic Kernel, and Haystack. Rasa could also be included, as it’s an open-source, high-code chatbot framework. While DialogueFlow is more low-code, Rasa has more coding involvement, so it might be relevant. I’ll search for information on Rasa and ensure to cite its official documentation. I'll avoid including Model Context Protocol, as it's more of an interface than a framework. Thought for 39 seconds Below is a curated list of high-code, developer-focused AI agent frameworks that require writing code to integrate, customize, and orchestrate agent workflows. Each of these libraries is widely adopted in production and research settings for building sophisticated single- and multi-agent systems. ## High-Code AI Agent Frameworks ### LangChain LangChain is a software framework in Python and JavaScript for building context-aware, reasoning applications with large language models, featuring chains, agents, memory modules, and tool integrations ([Wikipedia][1], [LangChain][2]). ### LangGraph LangGraph is a low-level orchestration framework built on LangChain that enables developers to construct stateful, graph-based multi-agent workflows with support for cycles, persistence, and human-in-the-loop approvals ([GitHub][3], [Langchain][4]). ### OpenAI Agents SDK The OpenAI Agents SDK is a Python library for orchestrating single-agent and multi-agent workflows, offering primitives such as Agents (LLMs with tools), Handoffs (delegation), Guardrails (input/output validation), and built-in tracing for visualization and debugging ([OpenAI GitHub][5], [GitHub][6]). ### Microsoft Semantic Kernel Semantic Kernel is an enterprise-ready, model-agnostic SDK for building and orchestrating AI agents and multi-agent systems across C#, .NET, Python, and Java, featuring plugin extensibility, planning capabilities, memory management, and seamless integration with OpenAI and other providers ([GitHub][7], [Microsoft Learn][8]). ### Haystack Agents Haystack Agents, part of the open-source Haystack framework by deepset, lets developers create AI agents capable of retrieval-augmented generation (RAG), tool use, and multi-step reasoning within customizable pipelines designed for production workloads ([Haystack Documentation][9], [Enterprise AI, Your Way | deepset][10]). ### Rasa Rasa is an open-source machine learning framework for building contextual AI assistants and chatbots in Python, providing natural language understanding, dialogue management, and flexible conversational pipelines for complex, stateful interactions ([rasa.com][11]). [1]: https://en.wikipedia.org/wiki/LangChain?utm_source=chatgpt.com "LangChain" [2]: https://www.langchain.com/?utm_source=chatgpt.com "LangChain" [3]: https://github.com/langchain-ai/langgraph?utm_source=chatgpt.com "langchain-ai/langgraph: Build resilient language agents as graphs." [4]: https://langchain-ai.github.io/langgraph/tutorials/introduction/?utm_source=chatgpt.com "LangGraph Quickstart - GitHub Pages" [5]: https://openai.github.io/openai-agents-python/?utm_source=chatgpt.com "OpenAI Agents SDK" [6]: https://github.com/openai/openai-agents-python?utm_source=chatgpt.com "openai/openai-agents-python: A lightweight, powerful ... - GitHub" [7]: https://github.com/microsoft/semantic-kernel?utm_source=chatgpt.com "microsoft/semantic-kernel: Integrate cutting-edge LLM ... - GitHub" [8]: https://learn.microsoft.com/en-us/semantic-kernel/frameworks/agent/?utm_source=chatgpt.com "Semantic Kernel Agent Framework | Microsoft Learn" [9]: https://docs.haystack.deepset.ai/docs/agents?utm_source=chatgpt.com "Agents - Haystack Documentation - Deepset" [10]: https://www.deepset.ai/products-and-services/haystack?utm_source=chatgpt.com "Build Custom AI Agents and Apps Faster | Haystack by deepset" [11]: https://rasa.com/docs/rasa/?utm_source=chatgpt.com "Introduction to Rasa Open Source & Rasa Pro" ## Conclusion For today’s AI Lab session, participants can choose any of these seven platforms to prototype an agent: **Crew AI** for multi-agent crews, **Langflow** for RAG workflows, **Botpress** for conversational bots, **n8n** for integrated pipelines, **Bubble** for end-to-end web apps, **Lindy** for self-learning workflows, or **Stack AI** for enterprise use. Each offers a unique balance of usability, extensibility, and integration—perfect for hands-on exploration and group debriefs at the end of the session. [1]: https://www.crewai.com/?utm_source=chatgpt.com "CrewAI" [2]: https://www.langflow.org/?utm_source=chatgpt.com "Langflow | Low-code AI builder for agentic and RAG applications" [3]: https://botpress.com/?utm_source=chatgpt.com "Botpress | The Complete AI Agent Platform" [4]: https://n8n.io/integrations/agent/?utm_source=chatgpt.com "AI Agent integrations | Workflow automation with n8n" [5]: https://bubble.io/blog/ai-agent/?utm_source=chatgpt.com "How to Build an AI Agent Without Code - Bubble" [6]: https://www.lindy.ai/?utm_source=chatgpt.com "Lindy — Meet Your AI Assistant" [7]: https://www.stack-ai.com/?utm_source=chatgpt.com "Stack AI: Build AI Agents with the Enterprise AI Platform" [8]: https://www.analyticsvidhya.com/blog/2024/11/build-your-first-agent-with-no-code-tools/?utm_source=chatgpt.com "Build LLM Agents on the Fly Without Code With CrewAI" [9]: https://www.datastax.com/products/langflow?utm_source=chatgpt.com "Langflow: Visual Low-Code AI App Builder for Agents and RAG" [10]: https://langflow.org/desktop?utm_source=chatgpt.com "Low-code AI builder for agentic and RAG applications - Langflow" [11]: https://botpress.com/blog/ai-agent-frameworks?utm_source=chatgpt.com "Top 5 Free AI Agent Frameworks - Botpress" [12]: https://blog.n8n.io/ai-agents/?utm_source=chatgpt.com "AI Agents Explained: From Theory to Practical Deployment - n8n Blog" [13]: https://bubble.io/ai?utm_source=chatgpt.com "Build your AI-powered app quickly - Bubble" [14]: https://www.lindy.ai/tools?utm_source=chatgpt.com "Custom AI Tools - Easy, Powerful, No-Code - Lindy" [15]: https://www.lindy.ai/blog/how-to-create-a-self-learning-ai-assistant---without-coding-4-steps?utm_source=chatgpt.com "How To Make an AI Assistant Without Coding: 9 Simple Steps - Lindy" [16]: https://mitsloanedtech.mit.edu/tools/stack-ai/?utm_source=chatgpt.com "Tool Overview: Stack AI" [17]: https://www.ycombinator.com/companies/stack-ai?utm_source=chatgpt.com "Stack AI: AI Agents for the Enterprise - Y Combinator"