Iris

Orchestrator for autonomous AI agent systems.

Living Documentation — Updated 24 Jun 2026

What is Iris?

Iris is not a chatbot. Iris is a persistent orchestrator that builds, manages, and heals specialized sub-agents. She runs 24/7 on an Azure VM, talks to you over Slack, sends scheduled reports by email, and commits every change to GitHub.

Every capability Iris adds is encapsulated as a skill — a reusable CLI tool stored in code. Skills are committed to GitHub before use, and every agent Iris spawns gets two environments (preview and production) provisioned via Terraform.

Architecture

User (Slack/Telegram)
    ↓
Iris Orchestrator (@iris-core/runtime)
    ├─ Memory (SQLite + GitHub repo)
    ├─ Skills (reusable CLI tools)
    ├─ Sub-agents (isolated containers)
    ├─ Events (cron / webhooks)
    └─ CDN (static assets, charts)

Skills Built

💰 aws-cost

Daily gross usage + MTD + forecast. Excludes credits. Auto-posts to Slack + email.

💰 azure-cost

Daily spend + MTD by service (INR + USD). Uses Consumption API. Same delivery.

📊 portfolio

Stock + mutual fund tracker. Live prices (yfinance + AMFI), P&L, alerts, charts.

📈 plotly-chart

Interactive charts (candlestick, line, area, OHLC). Publishes to CDN with public URL.

📊 google-finance

Stock quotes, charts, comparison charts, research reports from Google Finance.

🔍 klura

Browser-to-API capture and replay. Zero-token replay of logged-in web workflows.

🔑 get-secret

Cloud-portable secret retrieval. Env vars first, falls back to Azure Key Vault.

📧 send-email

Resend API integration for human escalation and scheduled reports.

Live Metrics

7
Skills Built
5
Daily Events
24/7
Uptime
$0
AWS Net Bill

Azure MTD (through 23 Jun): ₹80,680 ($843) | Forecast: ₹105,234

Hosted Services

What Iris Does Daily

All bills include yesterday's spend, MTD total, service breakdown, and end-of-month forecast.

For Investors / Partners

Iris is the infrastructure layer for a future where every company runs a fleet of specialized AI agents — not one monolithic LLM, but dozens of narrow sub-agents working in parallel. Each with two environments, self-healing capabilities, and full audit trails in GitHub.

The pattern is: discover → build → commit → test in preview → promote to production. Iris executes this loop for every new capability.

Talk to Rohit → View GitHub →