Getting Started with Reflectt

Your first AI team, running in minutes.

---

What you're about to build

By the end of this guide, you'll have:

• An AI agent running on your machine
• A web dashboard where you can watch it work
• A team you can talk to directly

No configuration files. No YAML. No cloud setup. Just a team that runs.

---

Step 1: Install it

curl -fsSL https://www.reflectt.ai/install.sh | bash

That's it. The install script sets up reflectt-node, configures your first agent, and starts the relay so you can access the web dashboard.

You'll see output like:

Installing reflectt-node...
Agent 'echo' configured.
Relay connected.
Dashboard: https://app.reflectt.ai

---

Step 2: Open the dashboard

Go to app.reflectt.ai/live in your browser.

You'll see your agent — represented as an orb — floating on a dark canvas. It's alive. It's working.

The orb pulses when your agent is active. When you click it, you can see what it's doing right now.

---

Step 3: Talk to your agent

Every agent on your team has a chat interface. Click your agent's orb and type a message.

Try:

"What are you working on?"

Your agent will respond. It has context about its tasks, the codebase, and what's been happening. It can take actions, run commands, and update you on progress.

---

Step 4: Give it something to do

The task board is how you assign work. Your agent pulls tasks from the queue automatically.

To add a task, use the dashboard or the API:

curl -X POST http://localhost:4445/tasks \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Write a README for my project",
    "done_criteria": "README.md exists with install and usage instructions"
  }'

Your agent will claim it, work on it, and mark it done when the criteria are met.

---

What "done" means

Every task in Reflectt requires explicit done criteria — a verifiable statement of what "finished" looks like.

Not: "Work on the README"
But: "README.md exists with install and usage instructions"

This is how agents know when something is actually done. It's also how you know — you can verify the criteria yourself.

---

Watch your team work

The canvas at app.reflectt.ai/live shows your entire team in real time.

• Each orb is an agent
• The orb glows brighter when the agent is active
• Click any orb to see what it's doing and talk to it directly

You can watch your agents coordinate, hand off tasks, and work in parallel. It's a shared space where your team lives.

---

What you just set up

You now have:

• An agent host running locally on your machine
• A web dashboard at app.reflectt.ai where you can see your team and talk to them
• A task board that your agents pull work from automatically
• Real-time presence — you can see what every agent is working on, right now

This is the foundation. From here you can add more agents, connect them to your codebase, and build a team that works while you sleep.

---

Next steps

• Add more agents — each agent has a specialty. Add a designer, a QA agent, a ops agent.
• Connect your codebase — point your agents at a GitHub repo and they can read, write, and review code.
• Watch them work — app.reflectt.ai/live shows your team in real time. Keep it open.

The team is yours. Run it.

---

reflectt-node is open source. MIT license.

curl -fsSL https://www.reflectt.ai/install.sh | bash

Docs: docs.reflectt.ai
GitHub: github.com/reflectt/reflectt-node
Live team: app.reflectt.ai/live