How Iris agents actually work
Most AI tools are reactive — you ask, they answer. Iris agents are different. They have memory, schedule themselves, use real tools, and follow up without prompting. Here's the six-step loop that makes that possible.
Tell your agent what you want
Plain English. Text it, chat in the dashboard, or speak. "Email my prospect list every Tuesday with a personal hook." That's the spec.
It learns who you are
The agent reads your identity, profile, and past conversations. Every interaction sharpens its memory — names, preferences, tone, deadlines. It's not a stateless chatbot.
It schedules itself
Routine work doesn't need you. The agent writes its own cron jobs — hourly, daily, on triggers — and quietly executes while you sleep.
It uses real tools
Gmail, Calendar, Slack, Stripe, Monday, Notion, your CRM — 100+ integrations through Composio MCP. Phone calls, SMS, image generation, web search — built-in.
It reports back, proactively
When something interesting happens — a deal moved, a deadline shifted, a customer replied — you get a text. No dashboard staring required.
It builds new skills as you go
If you ask for the same thing twice, the agent codifies it into a reusable skill. Over time it becomes more capable, not less. You're training your own AI.
From "I wish I had an assistant" to "my agent just sent it" — in under 2 minutes.
No code. No flowchart builder. No yaml. You describe the job once, and Iris handles the rest — finding the right tools, writing the right schedule, asking only the questions it can't answer itself.
Start your first agent