AI staff for a small company: they find sales, answer clients, write your posts and watch the money, on a machine you own. You approve what matters

1. Give each agent a job. Answer clients, send the daily report, or work on an issue.
2. It works in your team's tools. WhatsApp, Telegram, GitLab, GitHub and schedules, on a machine you own.
3. You check the result and decide. The browser dashboard shows what was done and what still needs you.

New here? Before you start lists what you need, how to install it, and how to check it works.

AgentX runs AI agents for a team. Connect a channel (a chat or work tool such as Telegram or GitLab), give an agent a job, and see what happened in the browser dashboard. A technical teammate installs it on a machine you control; operators can then use the browser for everyday work.

AgentX has two parts that run side by side:

• The daemon is the background service that receives messages, runs agents and runs schedules (work that starts at set times).
• The dashboard is the website you open in your browser to set things up and watch the work.

Where to start

• Try it without a model account: see the demo. It uses real AgentX daemons and scripted model replies, so it makes no paid model calls.
• Know what leaves your machine: Your data lists what goes to the model provider and other services, and how to keep it at home.
• Set up your own team:
  1. Install AgentX.
  2. Create your first agent.
  3. Connect Telegram.

The dashboard has eight main tabs: Live, Operations, Monitor, Approvals, People, Activity, Workflows, and Settings. Start with Monitor to understand what still needs a person and what agents can handle.

New words, such as node, mesh or schedule, are explained in plain terms in the glossary.

More ways to work with your agents

• In-page chat: ask about the dashboard view you are looking at.
• AgentX Voice: talk to any agent from anywhere on your Mac, and hear it answer in its own voice.
• AgentX Voice reference: the full reference for AgentX Voice, with talks between agents, live lessons, narration and automations.
• Terminal UI: use OpenCode as the conversation interface.
• Tailscale setup: connect machines privately.
• Agent-to-agent communication: send work to agents on other machines, or connect an outside tool that speaks A2A (an open standard for agents talking to each other).

Learn one step at a time

1. See it: follow the annotated workflow tour.
2. Try it: record a VS Code walkthrough.
3. Understand it: architecture and Jev decisions (small yes-or-no and pick-one questions that AgentX can hand to a cheaper, faster model).
4. Use it from a terminal: command reference.

Check it worked

Once AgentX is installed:

1. Terminal: run agentx daemon status. It prints Status: running and lists your agents.
2. Browser: open the dashboard (by default http://127.0.0.1:4202/live). The Live tab shows your agents.

If something is wrong

• You don't have AgentX yet: start with Before you start, then Install.
• The dashboard opens but agents don't answer: the daemon and the dashboard are separate. Check the daemon with agentx daemon status, then follow It's not answering.
• A word on these pages is unclear: look it up in the glossary.