About AgentSwarms
AgentSwarms is a browser based learning platform for building agentic AI systems. It pairs written lessons with runnable agents, notebooks and labs, so a learner reads a short explainer and then runs the same pattern live, forks it and modifies it. Topics span tool use, retrieval augmented generation, guardrails, planner and executor setups, multi agent swarms and text to SQL agents.
The platform is aimed at developers, analysts and technical teams who want to understand and prototype agents rather than only read about them. The curriculum is organized into eight tracks with more than 50 lessons, 97 notebooks, 21 slide decks and 25 build along labs. Templates provide working multi agent systems for operations, finance research and software release workflows, and these can be exported to a choice of frameworks. Agents can be given a voice through a speech to text, reasoning and speech output loop, embedded in other websites through an iframe widget, and published to a public community feed for others to remix. Controls such as human approval inboxes, traces and per agent budget caps are included. Users can bring their own keys for providers including OpenAI, Anthropic, Gemini, Bedrock, Azure, Grok and Qwen. A separate open source edition covers self hosting of agents, retrieval, evaluations and business intelligence.
The platform is free during its public beta, with no credit card required and a daily request cap on its hosted AI gateway. A Pro plan with higher limits and monthly gateway credits is listed as coming soon, and its price has not been announced.