Book demo

iris-labx

IRIS LABX

> booting Iris Labx research environment

We build post-trained models and autonomous agents for the AI first era.

LIMs turn human intention into action across tools, APIs, agents, and software systems.

Agentizer makes existing products executable by autonomous agents.

AgentNet is the capability layer for the World Wide Agents.

Contact

contact@irislabx.com

Hiring

open roles

Locations

Paris + San Francisco

Iris Labx

We are building the foundation for the AI-first world: the Netscape and Google Search moment for a new era.

LIM: Large Intention Model

A Large Intention Model is a post-trained model designed to understand what humans want to accomplish, map that intention to executable capabilities, and coordinate autonomous agents that act across digital and physical systems.

LLMs predict language. LIMs operationalize intention. Iris Labx is researching this shift from response generation to outcome execution for the AI first era.

Read the LIM explainer

Research

2025

Research / 2025

Intention Is All You Need for Autonomous Agents

A new paradigm for AI systems that understand and act on human intention, moving beyond simple prompt-response interactions.

Request access

The Digital World Needs a Reset

The internet was built for humans scrolling, clicking, and browsing. But the next generation of users won't be people.

They'll be trillions of autonomous AI agents.

From the World Wide Web to the World Wide Agents

The web was designed to display information. Not to execute intentions.

Today, AI agents still operate software like humans:

It's slow, fragile, and inefficient.

Iris Labx is building the infrastructure for the World Wide Agents (WWA), a new internet layer where software becomes executable by autonomous agents at planetary scale.

The Rise of Trillions of Agents

Soon, every person, company, device, workflow, robot, and machine will be powered by AI agents.

There won't be millions of agents. There will be trillions.

Agents that:

The current internet infrastructure cannot support that future.

What We're Building

Iris

An autonomous agent capable of understanding intentions, discovering capabilities, and executing real-world actions across digital systems.

Iris can:

Not just generating text. Executing outcomes.

Agentizer

Transform any software into an agent-ready system.

Agentizer enables companies to expose their products through:

Making software executable by AI agents instead of only usable by humans.

AgentNet

The discovery and distribution layer for autonomous capabilities. AgentNet indexes thousands of executable capabilities across MCP, WebMCP, WWA, CLI-native, A2A, and API-first ecosystems, enabling Iris and future autonomous agents to discover, reason, and execute actions across the real world.

Why It Matters

AI is evolving from conversation to execution.

The next trillion internet actions will not come from humans clicking interfaces. They will come from autonomous agents operating software directly.

But today's internet was never designed for trillions of autonomous systems interacting in real time.

The digital world needs a reset.

The Future Is Agent-First

The interface is changing:

Iris Labx is building the infrastructure layer powering this transition.

Build for the Agentic Era

Prepare your software for the next generation of users: trillions of autonomous AI agents.

Welcome to the World Wide Agents.