Automation with AI agents
Agentic automation connects workflows, interfaces and user interfaces – wherever fixed rules alone are no longer enough.
Classic automation follows fixed rules: if A happens, B follows. Agentic automation adds decisions that an agent makes within its mandate – for example, which next step makes sense or how to classify an unstructured input.
This category includes workflow platforms with agent building blocks as well as agents that operate desktop applications or entire operating systems (computer use).
Role in an AI-first business model
An AI-first operating model does not replace fixed rules and classic automation; it combines them with agents. Whatever can reliably be described as a rule stays a rule. Agents take over where judgement is needed.
The key question is which actions an agent prepares, recommends, executes with approval or performs autonomously within defined limits. This gradation makes automation manageable.
What to look for when choosing
- A combination of fixed rules and agent decisions
- Graded permissions per action
- Traceable states and repeatability
- Robustness against changes to user interfaces and APIs
13 tools in this category
DemoGPT
Open-source tool that turns a single prompt into a quick, runnable demo application. It applies a tree-of-thoughts approach to the LangChain documentation and is suited to rapid prototyping.
RestGPT
Research project for an LLM-based autonomous agent that controls real-world applications through RESTful APIs. Relevant for anyone who wants agents to carry out tasks via existing web interfaces.
XAgent
Autonomous LLM agent from OpenBMB designed to break down and work through complex tasks on its own. Useful for experimenting with autonomous task solving by agents.
LLM Agents
Lean library for building agents that are controlled by a large language model. It is a good way to understand the basic principle of agents that call tools.
uAgents
Lightweight framework from Fetch.ai for creating decentralized agents quickly. Intended for scenarios in which several independent agents need to communicate with each other.
Maestro
Framework in which Claude Opus acts as an orchestrator, breaking a task down and delegating it to subagents. A clear example of the orchestrator-subagent pattern for automating multi-step tasks.
llama-agents
Async-first framework from the LlamaIndex ecosystem for building, iterating on and running multi-agent systems in production. It covers inter-agent communication, distributed tool execution and human-in-the-loop, among other things.
Agno
Open-source platform for building, running and managing agent platforms. It is aimed at teams that want not only to develop agents but also to keep them in ongoing operation.
AgentK
Experimental, modular agent system designed to evolve itself by creating new agents and tools as needed. More of a research and learning project than a production tool.
ADAS
Research project on the Automated Design of Agentic Systems. It explores how agent architectures can be searched for and improved automatically instead of being designed by hand.
Giselle
Builder for agentic workflows that lets you assemble AI-driven processes. Suited to teams that want to set up workflow automation with AI agents without coding everything from scratch.
RedAmon
Self-hosted AI agent for penetration testing built on LangGraph. It maps the attack surface as a graph, tests it from a Kali sandbox with human approval gates and proposes fixes as pull requests.
OpenConnector
Open-source connector gateway through which agents can discover and execute actions in SaaS applications via MCP, HTTP or OpenAPI. It manages OAuth and API key credentials and issues scoped runtime tokens.
From tool to outcome
In AI-first coaching you clarify which outcomes AI agents should be accountable for in your go-to-market and within which framework. REVAGO then designs, builds and runs your system. You pay for results – not for activity.
Key concepts
Other categories
A neutral overview, not a recommendation of individual providers. REVAGO is not affiliated with the listed projects. Descriptions are based on the projects’ public information; features and licences may change. Selection based on awesome-agents (kyrolabs), As of 19 October 2026.