Laguna by Poolside

Product Introduction

1. Definition: Laguna is a series of foundation models (specifically models named XS.2 and M.1) developed by Poolside, a frontier AI lab. These models are purpose-built for long-horizon work, with a primary application in creating sophisticated software agents and automating complex software engineering tasks. They represent the core intelligence layer of Poolside's enterprise AI platform.
2. Core Value Proposition: Laguna exists to transform enterprise software development by driving measurable business outcomes, not just token generation. Its core proposition is to provide foundation models for AI agents that can operate securely within an organization's boundary, delivering intelligence across the development ecosystem—from IDE to terminal, agents to custom applications—to create abundance through artificial general intelligence (AGI).

Main Features

1. Foundation Models for Agents (Laguna XS.2, M.1): These are the core large language models optimized for tasks requiring multi-step reasoning, planning, and execution over extended periods. How it works: Trained with methodologies like reinforcement learning to understand the "why" behind software decisions, not just the "what," the models enable agents to plan, use tools, and execute tasks in sandboxed environments. They are designed to mirror human reasoning in software development.
2. Agents and Multi-Agent Orchestration: Poolside provides a framework for building single and multi-agent systems. How it works: These systems leverage the Laguna foundation models to plan complex tasks, utilize integrated tools, and execute actions. All agent activities are governed by configurable policies and include full end-to-end traces for auditability, making them suitable for governed enterprise workflows.
3. Developer Surfaces and Integration: The platform offers multiple interfaces to embed AI capabilities into existing developer workflows. This includes agents, terminal UIs (TUI), IDE extensions, binary tools, and workflow integrations. These surfaces ensure the AI intelligence moves with the engineers where they work, enhancing productivity without requiring a context switch.
4. Data and Knowledge Connectors: The system provides secure connectors to an organization's internal data repositories. This includes repositories (e.g., Git), databases, data warehouses, and private corpora. The connectors operate within strict security boundaries, allowing models to leverage internal knowledge without data ever leaving the enterprise control.
5. Enterprise-Grade Deployment and Governance: Laguna models and the entire Poolside system are designed for deployment inside an organization's security boundary—on-premises, in a private VPC, or on workstations (for defense use cases). It features role-based access control (RBAC) for both humans and agents, risk controls, and auditability co-designed to meet the requirements of enterprise review boards and CISOs.

Problems Solved

1. Pain Point: Enterprises struggle to harness the potential of AI for software development due to concerns over data security, model accuracy for complex tasks, governance, and integrating AI into heterogeneous, legacy, or air-gapped environments. Traditional AI code assistants often operate as isolated tools without systemic integration or enterprise controls.
2. Target Audience: Enterprise Technology Leadership (CTOs, VPs of Engineering), Software Architects, Development Team Leads, CISOs (Chief Information Security Officers), and Governance & Compliance Officers in large organizations, particularly in high-consequence industries like finance, defense, and critical infrastructure.
3. Use Cases: Automating multi-step software refactoring and migration tasks across complex codebases; building custom software agents that interact with internal APIs and data stores to perform debugging or deployment tasks; deploying AI-powered development tools in air-gapped or highly regulated networks; and conducting automated, auditable code analysis that adheres to internal policies.

Unique Advantages

1. Differentiation: Unlike generic LLM API providers or cloud-based AI coding assistants, Poolside differentiates with a "Your data. Your models. Your future." approach. It focuses on outcome ownership rather than token delivery, jointly taking responsibility for business impact. Models are deployed inside the enterprise security boundary, and the system is built for complex, heterogeneous environments including multi-cloud and legacy systems, avoiding a "rip-and-replace" mandate.
2. Key Innovation: The key innovation is the co-design of foundation models and an enterprise deployment system for long-horizon, high-consequence software work. This includes building models specifically for the reasoning demands of software engineering and engineering the entire platform—with governance, connectors, and orchestration—to function as a secure, integrated AI system of record for the development lifecycle, not just a model API.

Frequently Asked Questions (FAQ)

1. How does Poolside Laguna ensure our proprietary code and data remain secure? Poolside Laguna is designed for deployment inside your security boundary, whether on-premises, in your VPC, or on workstations. Your data never leaves your control. The system includes enterprise-grade governance with role-based access control (RBAC) for both humans and AI agents, alongside full audit trails and risk controls, meeting stringent CISO and review board requirements.

2. Is Laguna suitable for integrating with our existing legacy systems and development tools? Yes. Poolside is explicitly built for complex, heterogeneous environments. It provides data and knowledge connectors to your existing repositories, databases, and data warehouses. Its developer surfaces (IDE extensions, TUI, agents) are designed to integrate with your engineers' existing workflows across multi-cloud and legacy systems without requiring a disruptive rip-and-replace of your current stack.

3. What makes Laguna different from using a general-purpose LLM API like GPT-4 for coding tasks? Laguna is a specialized foundation model for long-horizon software work, built and battle-hardened for enterprise-scale, high-consequence applications. While general-purpose APIs are versatile, Laguna is optimized for multi-step reasoning, planning, and execution within a governed, auditable agent framework. Furthermore, it is delivered as a complete enterprise system with security, deployment options, and outcome-focused support that raw API access does not provide.