Nelly

Nelly is a no-code AI agent platform delivered as a desktop application that lets users build, test, and run autonomous agents using natural language instructions. Agents and their data are stored locally on the user's machine, while the LLM running decision-making is hosted in the cloud — Nelly currently uses Anthropic's Claude Sonnet 4 by default. Paid users can connect their own provider account from OpenAI, Anthropic, OpenRouter, or LM Studio to bypass platform message limits.

The platform supports multi-agent architectures where a parent agent breaks down complex tasks and delegates them to specialized sub-agents (each delegated message counts toward the user's quota). Agents can be given tools — currently a web browser and database access — to act on the world. Nelly is designed to evolve toward fully autonomous teams that require only strategic direction, with a marketplace planned for publishing and monetizing agents and tools.

• No-Code Agent Studio — Build agents by writing instructions in natural language; no programming required. Format with headings, lists, and other content blocks.
• Sub-Agent Delegation — Break complex tasks into smaller pieces and delegate to specialized sub-agents that operate as a coordinated team.
• Local-First Storage — All agent definitions, conversation threads, and data stay on the user's computer; only conversation context is sent to the LLM provider during execution.
• Bring Your Own LLM — Pro and Pro+ subscribers can connect OpenAI, Anthropic, OpenRouter, or LM Studio accounts directly to bypass the platform's message limits.
• Agent Toolkits — Give agents access to tools like a web browser and a database so they can act on the world rather than only generate text.
• Threaded Conversations — Organize chats with agents in threads to keep context tidy and improve agent performance.
• Departments and Drag-and-Drop Organization — Group agents into departments and rearrange them visually to build a structured AI team.
• Activity Insights — Step through agent activity, view tool calls, and see browser screenshots to debug behavior.
• Marketplace (Coming Soon) — Publish AI agents and tools to share or monetize them when the marketplace launches.