--- title: "Agent builders · visual and low-code surfaces · Agent Fieldbook" url: https://agentfieldbook.org/software/builders/ description: "The low-code side of the agent stack: OpenAI Agent Builder (deprecating), Microsoft Copilot Studio, Palantir AIP Agent Studio, Google Gemini Enterprise Agent Platform, Flowise and Rivet." section: "Software · Builders" source: Agent Fieldbook — generated from the published page --- # Agent builders The drag-drop canvases and low-code studios that let analysts, PMs and integrators compose agents against runtimes engineers built. ## Agent builders · visual and low-code surfaces Live data ## visual canvas · 3 ### Flowise ## FlowiseAI Open-source drag-and-drop UI for building LLM agents and RAG workflows. Wraps LangChain and LlamaIndex primitives as visual nodes so multi-agent flows, single-agent chatbots with RAG, and chat assistants can be composed by non-programmers. 100+ LLM providers, vector databases, and embeddings supported. Cloud or on-prem; embed via widget, API, or SDK. The OSS reference point for the visual-builder pattern outside vendor platforms. ## Runtime ## Deploys ## License - [↗ flowiseai.com](https://flowiseai.com/) ### Rivet ## Ironclad Open-source visual programming environment for building, debugging and deploying LLM-powered agents. Node-graph editor; graphs stored as YAML so they version and review like code. Ships with its own remote executor allowing real-time observation of agent execution from inside the host application. Built by Ironclad and used to power their Contract AI assistant. Embed into Node.js or TypeScript apps via the Rivet SDK — the code-adjacent visual builder for engineering teams that want the diagram but not the vendor lock-in. - [↗ rivet.ironcladapp.com](https://rivet.ironcladapp.com/) ### OpenAI Agent Builder ## OpenAI Drag-and-drop visual canvas for building multi-step agent workflows. Typed inputs and outputs on every node, live-data preview during authoring, templates for common patterns. Exports Agents SDK code for self-hosting or embeds via ChatKit. OpenAI has announced deprecation on 2026-11-30 and is steering new work to ChatKit (the embed surface) plus the Agents SDK (the code path) instead of the canvas — a notable step back from visual authoring by a top-tier vendor while Microsoft, Google, Palantir and Salesforce lean further in. - [↗ developers.openai.com](https://developers.openai.com/api/docs/guides/agent-builder) ## low-code studio · 2 ### Microsoft Copilot Studio ## Microsoft Graphical low-code builder for agents and agent flows. Compose topics (branching conversation threads), knowledge sources, tools and triggers via a visual designer; layer NLU on top so the agent maps free-text intents to the right topic. Two flow formats: classic agent flows (Power Automate style) and the new workflows (revamped visual designer with improved testing). Prebuilt and custom connectors reach enterprise data. Publish to Teams, web, mobile, or any channel supported by Azure Bot Service; also usable as a Microsoft 365 Copilot extension. - [↗ learn.microsoft.com](https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio) ### Palantir AIP Agent Studio ## Palantir Technologies Visual low-code interface for constructing agents that are anchored in Palantir's Ontology. Developers and analysts specify prompts, bind the agent's allowed Action Types and Object Types, and expose the agent through the Agent Engine SDK. Agents can be composed into fleets via Mindkit and re-exposed to external frameworks (Google ADK, Microsoft Agent Framework, Claude Desktop) through Ontology MCP — same governance guardrails inside and outside the Palantir perimeter. - [↗ palantir.com](https://palantir.com/docs/foundry/aip-agent-studio/overview/) ## hybrid low-code + code · 1 ### Google Gemini Enterprise Agent Platform (formerly Vertex AI Agent Builder) ## Google Cloud Enterprise agent platform for building, deploying, governing and optimising production agents. Combines the Agent Development Kit (ADK, the open-source code-first framework), Conversational Agents (formerly Dialogflow, for conversational surfaces), Vertex AI Search (grounding), and access to 200+ Google and third-party models. Rebranded from Vertex AI Agent Builder in 2026 under the Gemini Enterprise umbrella. Hybrid: low-code Conversational Agents alongside code-first ADK, so the same team can move between visual authoring and programmatic control. - [↗ cloud.google.com](https://cloud.google.com/products/agent-builder) ## Field notes The visual and low-code side of the agent stack Most of the market is building managed and headless agents for engineers. A different group wants to build their own: growth marketers, product managers, analysts, semi-technical people with a specific job to do and a reason to keep the cost down. Agent builders exist for them. The form is a drag-and-drop visual studio. Microsoft Copilot Studio, Palantir's AIP Agent Studio, Google's Gemini Enterprise platform, Flowise and Rivet all take it, as did OpenAI's Agent Builder until its deprecation date in November 2026. None of this is new: HubSpot and Braze made the visual workflow canvas standard in SaaS years ago, and Braze built much of its pitch on branching a customer journey on screen. What is new is pointing that familiar pattern at agents, where seeing the workflow laid out is most of how a non-technical user comes to understand what an agent is and how its parts connect. - [Brandon Chaplin](https://www.linkedin.com/in/brandon-chaplin-digital-marketing-strategist) ## Common questions ### How is an agent builder different from a framework? A framework is code: you compose the agent in Python or TypeScript. A builder is a canvas: you drag nodes into place and configure them in a browser. The runtime underneath is often the same framework, so the real difference is who can author, an engineer or an analyst. ### What is a low-code AI agent builder? A tool for assembling an agent by connecting blocks on a canvas instead of writing code. You pick a model, add tools, set the steps and publish. Microsoft Copilot Studio, Google Gemini Enterprise and Flowise are examples. [(Copilot Studio)](https://learn.microsoft.com/en-us/microsoft-copilot-studio/) ### Which agent builders are open source? Flowise and Rivet. Flowise is Apache 2.0 and wraps LangChain and LlamaIndex as drag-and-drop nodes. Rivet is MIT, built by Ironclad, and stores its graphs as YAML so they can go through normal code review. [(Flowise)](https://github.com/FlowiseAI/Flowise) ### Can I move an agent from one builder to another? Not directly. Each vendor runs its own runtime and its own graph format, so there is no export another canvas will read. What does travel is the tools layer: an MCP server built for one agent works with any other that speaks MCP [(Model Context Protocol)](https://modelcontextprotocol.io/). ### Do you need to write code to build an AI agent? Not for a working one. A low-code builder covers most single-agent workflows: a model, a few tools, a prompt and a trigger. Code becomes necessary when you need custom control flow, version control, tests, or your own deployment and scaling. ## Next in the learning path - [Frameworks The code-first orchestration runtimes underneath these builders](https://agentfieldbook.org/software/frameworks/) - [MCP How agents authored in different builders talk to each other](https://agentfieldbook.org/software/mcp/) - [Deployment modalities Where the built agent runs, headless, caged, cloud, local](https://agentfieldbook.org/catalog/modalities/) ## Entries _6 entries listed on this page._ | name | description | applicationCategory | operatingSystem | url | | --- | --- | --- | --- | --- | | OpenAI Agent Builder | Drag-and-drop visual canvas for building multi-step agent workflows. Typed inputs and outputs on every node, live-data preview during authoring, templates for common patterns. Exports Agents SDK code for self-hosting or embeds via ChatKit. OpenAI has announced deprecation on 2026-11-30 and is steering new work to ChatKit (the embed surface) plus the Agents SDK (the code path) instead of the canvas — a notable step back from visual authoring by a top-tier vendor while Microsoft, Google, Palantir and Salesforce lean further in. | BusinessApplication | Web | https://developers.openai.com/api/docs/guides/agent-builder | | Microsoft Copilot Studio | Graphical low-code builder for agents and agent flows. Compose topics (branching conversation threads), knowledge sources, tools and triggers via a visual designer; layer NLU on top so the agent maps free-text intents to the right topic. Two flow formats: classic agent flows (Power Automate style) and the new workflows (revamped visual designer with improved testing). Prebuilt and custom connectors reach enterprise data. Publish to Teams, web, mobile, or any channel supported by Azure Bot Service; also usable as a Microsoft 365 Copilot extension. | BusinessApplication | Web | https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio | | Palantir AIP Agent Studio | Visual low-code interface for constructing agents that are anchored in Palantir's Ontology. Developers and analysts specify prompts, bind the agent's allowed Action Types and Object Types, and expose the agent through the Agent Engine SDK. Agents can be composed into fleets via Mindkit and re-exposed to external frameworks (Google ADK, Microsoft Agent Framework, Claude Desktop) through Ontology MCP — same governance guardrails inside and outside the Palantir perimeter. | BusinessApplication | Web | https://palantir.com/docs/foundry/aip-agent-studio/overview/ | | Google Gemini Enterprise Agent Platform (formerly Vertex AI Agent Builder) | Enterprise agent platform for building, deploying, governing and optimising production agents. Combines the Agent Development Kit (ADK, the open-source code-first framework), Conversational Agents (formerly Dialogflow, for conversational surfaces), Vertex AI Search (grounding), and access to 200+ Google and third-party models. Rebranded from Vertex AI Agent Builder in 2026 under the Gemini Enterprise umbrella. Hybrid: low-code Conversational Agents alongside code-first ADK, so the same team can move between visual authoring and programmatic control. | BusinessApplication | Web | https://cloud.google.com/products/agent-builder | | Flowise | Open-source drag-and-drop UI for building LLM agents and RAG workflows. Wraps LangChain and LlamaIndex primitives as visual nodes so multi-agent flows, single-agent chatbots with RAG, and chat assistants can be composed by non-programmers. 100+ LLM providers, vector databases, and embeddings supported. Cloud or on-prem; embed via widget, API, or SDK. The OSS reference point for the visual-builder pattern outside vendor platforms. | BusinessApplication | Web | https://flowiseai.com/ | | Rivet | Open-source visual programming environment for building, debugging and deploying LLM-powered agents. Node-graph editor; graphs stored as YAML so they version and review like code. Ships with its own remote executor allowing real-time observation of agent execution from inside the host application. Built by Ironclad and used to power their Contract AI assistant. Embed into Node.js or TypeScript apps via the Rivet SDK — the code-adjacent visual builder for engineering teams that want the diagram but not the vendor lock-in. | BusinessApplication | Web | https://rivet.ironcladapp.com/ |