What is an AI agent?
An AI agent is an AI model, such as ChatGPT, Claude or Grok, that can take actions with tools you give it, like reading a message and sending a reply, rather than only chatting.
In plain words
Most people first meet AI as a chat window: you type a question and it writes an answer. An AI agent goes one step further. It is given tools, and it decides when to use them to get a job done. A tool might read your latest messages, look up a booking or send a reply.
For a small business, that difference matters. A chat window can draft a reply that you then copy and paste. An agent can read the customer's message on your own account, check the notes you gave it about prices and opening hours, and answer in the same conversation while you are busy with a client.
An agent is only as good as its instructions and its tools. You tell it what your business does, how you like to sound, and when it should stop and pass the conversation to you. The tools decide what it is able to do at all. An agent without a messaging tool cannot send a message, however it is asked.
Agents still make mistakes, so it is sensible to start small: let it answer common questions first, read what it sends, and widen what it handles as you gain confidence.
How it works in Spaces
Spaces gives the AI agent you already use the tools to read and reply on your own WhatsApp, Instagram, LinkedIn, Messenger, X and email accounts. It works with Claude Code, Claude, Cursor, Codex, ChatGPT, Grok, OpenClaw and Hermes.
Spaces keeps every account inside safe sending limits on its own side, so the agent cannot send faster than Spaces allows, whatever it is told. Some messages are always refused, such as the same cold message sent to many people, or anything to someone who said stop.
Related terms
- MCP: MCP, the Model Context Protocol, is an open standard that lets an AI agent connect to an outside service and use its tools, a bit like a universal plug between AI apps and the services they work with.
- MCP server: An MCP server is the web address a service publishes so AI agents can find and use its tools through the Model Context Protocol.
- API key: An API key is a long secret code that proves to a service that a request comes from you, so it works like a password for software rather than for people.
- Connected account: A connected account is one of your own messaging accounts, such as your WhatsApp number or Instagram profile, that you have linked to a service so an AI agent can read and reply to its conversations.
- Hand-over: A hand-over is when an AI agent stops replying in a conversation and passes it to a person, because the customer needs judgement, care or a decision the AI should not make.
FAQ
Is an AI agent the same as a chatbot?
Not quite. A traditional chatbot follows a script you build. An AI agent understands the message, decides what to do and uses tools to do it, such as reading a conversation and sending a reply.
Which AI agents work with Spaces?
Claude Code, Claude, Cursor, Codex, ChatGPT, Grok, OpenClaw and Hermes all connect to Spaces through its MCP server.
Can an AI agent send messages from my own account?
Yes, once you connect the account to Spaces. Customers see replies from the same number, profile or email address they already know.
Can I stop the agent at any time?
Yes. You can pause all sending for the workspace, pause one API key, or hand a single conversation to a person so the agent stays out of it.
Will the agent always get things right?
No AI is perfect. Give it clear notes about your business, tell it when to hand a chat to you, and check its replies while you get started.
Useful next: AI agents that work with Spaces, What to tell your AI agent, Account safety. Every term: the glossary.