AI agents wiki · REVAGO

What are AI agents?

An AI agent is a system that pursues a defined goal and selects and executes steps on its own within approved limits. Unlike a chatbot that answers a question, an agent plans, calls tools and APIs, checks intermediate results and keeps working until the goal is reached or a person takes over.

Agentic workflows combine language models with fixed rules, classic automation and human decisions. How much an agent may do on its own is not a technical decision but a business one.

AI agents in AI-first and AI-native business models

The tools in this wiki are means to an end. Their value only emerges in the business model – either when an existing company redesigns its processes from the outcome (AI-first), or when an offering is built on agents from the start (AI-native).

AI-first business model: An AI-first business model designs processes from the desired outcome for AI agents and people – instead of keeping existing workflows unchanged and merely speeding them up with AI tools.
AI-native business model: An AI-native business model is built on AI agents from the start: agents carry the core of value creation, and the offering would not exist in this form without them.

The categories

The selection of categories and projects follows the open-source list awesome-agents, complemented by LangGraph and DeepAgents. Each category explains what the tools do, what role they play in an AI-first business model and what to look for when choosing.

AI agent frameworks62 tools: Frameworks provide the scaffolding that lets AI agents plan, call tools, keep state and work together – from LangChain, LangGraph and DeepAgents to CrewAI, AutoGen and Hermes Agent.
Testing and evaluating AI agents8 tools: Tools for testing, evaluation, tracing and observability show whether an AI agent does what it should – before and while it operates.
AI agents for software development56 tools: Coding agents write, review and change code, work on issues and operate in the terminal, editor or repository – from assistants to largely autonomous development agents.
Research agents13 tools: Research agents search sources, condense information and write reports – from market research to scientific literature reviews.
Conversational agents and general agents18 tools: Conversational and general-purpose AI agents work via chat, messengers or the desktop and take on changing tasks – from personal assistants to generalists with tool access.
Memory and knowledge management for AI agents13 tools: Memory systems, knowledge graphs and RAG tools give AI agents a memory – so they know customers, cases and company knowledge beyond a single conversation.
Automation with AI agents13 tools: Agentic automation connects workflows, interfaces and user interfaces – wherever fixed rules alone are no longer enough.
Browser agents7 tools: Browser agents operate websites like a person: they navigate, read, click and fill in forms – even where no API exists.

In focus

Four projects people ask about most, each with its own page:

LangChain: LangChain is an open-source framework for AI agents and LLM applications. It connects language models to tools, data and systems through a unified interface – and is the entry point to the ecosystem of LangGraph, DeepAgents and LangSmith.
LangGraph: LangGraph is a low-level framework for orchestrating long-running, stateful AI agents. Workflows are modelled as a graph – with durable execution, human-in-the-loop and memory.
DeepAgents: DeepAgents (Deep Agents) is an open-source agent harness by LangChain: an opinionated agent that runs out of the box and already includes planning, sub-agents, a filesystem, context management and approvals.
Nous Hermes and Hermes Agent: Nous Research is behind two things that are often confused: the Hermes family of open-weight models and Hermes Agent, a self-improving open-source agent that works with any model.

Using agent tools effectively

Whether a framework or agent is effective in a company is rarely decided by technology alone. Five questions belong at the start:

Outcome first: Which business outcome should the agent be accountable for – and how do you measure it? Activity is not an outcome.
Mandate and accountability: Who sets goals, who decides, who intervenes? An agent needs an authorised mandate, not a blank cheque.
Graded permissions: Which actions does the agent prepare, which does it recommend, which does it execute with approval, which does it perform autonomously within defined limits?
Evaluation: How do you check quality and impact before an agent gets more room to act – and while it works?
Operation: Who maintains, monitors and improves the system on an ongoing basis? An agent is never finished.

AI-first coaching →Check your REVAGO potential

Glossary: AI agent terms

The key terms briefly explained.

AI agent: A system that pursues a defined goal and selects and executes steps within approved limits – using language models, tools, rules and human decisions.
Agentic AI: Umbrella term for AI systems that do not just answer but plan and act. In sales, this is called agentic sales.
AI-first: Designing processes from the desired outcome for agents and people, instead of merely speeding up existing workflows with AI tools.
AI-native: A business model built on AI agents from the start that would not exist in this form without them.
AI-first operating model: An operating model in which agents, fixed rules, classic automation and human decision points work together.
Human-in-the-loop (HITL): Deliberately designed human decision, review and intervention points in an agentic workflow.
Multi-agent system: Several specialised agents share a task, for example through delegation to sub-agents or through roles in a team.
Agent harness: A preconfigured agent with a defined way of working (planning, context, tools) that can be extended – such as DeepAgents.
Orchestration: Control of order, branches and handovers in an agentic workflow – for example as a graph in LangGraph.
Tool calling (function calling): The ability of a language model to call tools or APIs in a structured way and process their results.
Model Context Protocol (MCP): An open standard through which agents can access tools, data sources and applications.
Retrieval-augmented generation (RAG): Before answering, the agent searches documents or databases and relies on what it finds.
Agent memory: Short- and long-term memory that lets an agent retain information beyond a single task or session.
Evaluation (evals): Systematic assessment of agent behaviour – of individual answers as well as entire workflows.
Tracing and observability: Traceable logs of decisions, tool calls and costs during operation.
Sandbox: An isolated execution environment in which an agent runs code or commands without endangering other systems.
Computer use and browser agents: Agents that operate the user interfaces of desktop applications or websites like a person.
Open-weight model: A language model whose weights are published and which can be self-hosted – such as the Hermes models by Nous Research.
Prompt injection: Hidden instructions in content intended to lead an agent into unwanted actions. Limits are therefore enforced at the tool level.

Frequently asked questions about AI agents

What is an AI agent?

An AI agent is a system that pursues a defined goal and selects and executes steps on its own within approved limits. It plans, calls tools and APIs, checks intermediate results and keeps working until the goal is reached or a person takes over.

Which framework is suitable for AI agents?

It depends on the task. LangChain suits agents with many integrations, LangGraph controllable, stateful workflows with human-in-the-loop, DeepAgents long-running tasks with planning and sub-agents. CrewAI, AutoGen or LlamaIndex set other priorities. More important than the choice is clarifying outcome, mandate, approvals and operation beforehand.

How do you use AI agents effectively in an AI-first business model?

Start with the outcome, not the tool: define which business outcome an agent is accountable for, who grants the mandate and which actions it prepares, recommends, executes with approval or performs autonomously. Then follow evaluation before and during use and ongoing operation that maintains and improves the agent.

What is the difference between AI-first and AI-native?

AI-first describes an existing company that redesigns its processes from the outcome for agents and people. AI-native describes a business model built on AI agents from the start that would not exist in this form without them.

Does REVAGO use the tools in this wiki?

The wiki is a neutral overview. REVAGO is not affiliated with any of the listed projects. Which technology is used in a REVAGO engagement depends on your goal, your mandate and your systems.

All tools from A to Z

Each project links to its entry in the respective category. 190 tools.

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.

Check your REVAGO potential →AI-first coaching

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.