How to Chat With Your AI Agents on Macha
There are three ways to work with a Macha agent: chat with it from the main window, chat with it from inside the agent, or let it run on its own from a trigger. Chat is the hands-on mode — it's how you test an agent, run one-off tasks, and work alongside it like a copilot, all while staying in control of anything it changes.
Watch it
Three ways to start a chat
You can reach an agent's chat from a few places, whichever fits your flow:
- From the main window. Hit the button to see a list of your available agents and pick one (you can also click the agent icon to open the same list).
- From the agent itself. Open an agent from the left menu, then use the chat bar at the top. The agent is already selected, and you'll get a few helper prompts to get started.
Either way you land in the same place: a conversation with the agent you chose.
What chat is for
Chat isn't just a demo mode — it's a genuinely useful way to work:
- Test an agent. Before you let an agent run autonomously, chat with it on real examples and watch what it does and what it would change. (More in how to test an agent.)
- One-off tasks. Need something done once — summarize this ticket, look up that order, draft a reply? Just ask the agent in chat instead of building automation for it.
- Work side by side. Use the agent as a copilot while you handle a ticket — let it pull data, draft, and suggest while you stay in the driver's seat.
You're always in control
The key thing about chat: write actions ask for confirmation. When an agent is about to edit, update, or delete something — change a ticket's fields, post a public reply — it pauses and asks you to confirm first. Nothing changes without your okay.
This is the big difference from autonomous mode, where an agent running from a trigger acts on its own with no confirmation step. Chat keeps a human in the loop by design, which is exactly why it's the right place to build trust in an agent before you let it run free. (See autonomous mode vs. interactive chat for the full comparison.)
How chat fits the bigger picture
A typical agent lifecycle uses both modes:
- Build the agent.
- Chat with it to test and refine — watch its decisions, approve or reject its actions, tighten the instructions.
- Add a trigger and let it run autonomously once you trust it.
- Come back to chat any time for ad-hoc tasks or to debug behavior.
So chat isn't a beginner mode you grow out of — it's the hands-on counterpart to autonomous running, useful throughout an agent's life.
Chat vs. automate: which to use when
The two modes aren't competitors — they're for different moments:
| You want to… | Use |
|---|---|
| Try an agent before trusting it | Chat — test it, approve its actions |
| Do something just once | Chat — just ask |
| Work a ticket alongside the agent | Chat — copilot style |
| Handle every new ticket hands-free | A trigger (autonomous) |
| Run a job on a schedule | A scheduled trigger |
The rule of thumb: chat when you want a human in the loop; use a trigger when you want the work to happen without you. Most agents start their life in chat — where you test and refine them with confirmation on every write — and graduate to a trigger once you trust them. And even after they're autonomous, chat stays useful for one-off tasks and for debugging when something looks off.
Frequently asked questions
How do I chat with an agent? Hit the button (or agent icon) to pick from your agent list, or open the agent from the left menu and use the chat bar — it'll be pre-selected with helper prompts.
What's chat good for? Testing agents, one-off tasks, and working alongside an agent as a copilot.
Will the agent change things while I chat? Only with your confirmation — write actions pause and ask before they run.
How is chat different from autonomous mode? Chat keeps a human in the loop (confirmation on writes); autonomous (trigger-fired) runs with no confirmation.
Do I have to choose between chat and automation? No — most agents are tested in chat, then run autonomously on a trigger, and you can still chat with them any time.
The bottom line
Chat is the hands-on way to work with your agents — start one from the main window or the agent itself, use it to test, do one-off tasks, or copilot a ticket, and rely on confirmation-before-write to stay in control. It's where you build trust in an agent before turning it loose on a trigger.
Try it: build an agent and chat with it on a real ticket. 7-day free trial, no credit card required. Start free.
