Last updated 06-28-2026
Category:
Reviews:
Join thousands of AI enthusiasts in the World of AI!
Poolside
Poolside is a frontier AI lab building foundation models and agent systems for software engineering. The company trains models like Laguna M.1 and Laguna XS.2 in-house, then ships them through developer surfaces from terminal CLI to IDE extensions and a governance Console for platform teams.
The platform is built for deployment inside a customer security boundary. Foundation models arrive as full weights you run on-prem, in your VPC, or on air-gapped hardware. Agents execute in sandboxes with fine-grained permissions, checkpointed work layered on your version control, and full trajectory logs administrators can search and export.
Poolside targets enterprises and regulated sectors where generic cloud coding assistants are not an option. Forward Deployed Research Engineers embed with customer teams to deliver production agent workflows rather than handing off models alone. The stack supports multi-agent orchestration, OpenAI-compatible APIs, and third-party model providers through the same Console.
Laguna M.1 and XS.2 MoE models trained in-house for agentic coding
Run agents from CLI, IDE, web, or headless inside CI/CD pipelines
Sandboxed execution with checkpoints layered on your version control
OpenAI-compatible API so you keep your existing SDK patterns
Deploy on-prem, VPC, or air-gapped with full model weights delivered
Console traces every agent action with searchable, exportable trajectories
Full platform deployable inside your security boundary with no data egress required.
Laguna XS.2 ships as open-weight under Apache 2.0 for community evaluation and fine-tuning.
Every agent session produces searchable, exportable trajectories for compliance review.
OpenAI-compatible API and ACP support let teams adopt Poolside without replacing existing tooling.
No public self-serve pricing; enterprise deployment starts with a sales conversation.
Laguna M.1 weights require an institutional research request rather than open download.
The full stack is oriented toward large enterprises and regulated environments, not hobby projects.
Does Poolside have public pricing?
Poolside does not publish self-serve pricing on its website. The enterprise platform is sold through sales conversations, and the contact page invites teams in financial services, defense, and industrial sectors to talk with Poolside about deployment.
Can I try Poolside models for free?
Yes, for a limited time. Poolside lists Laguna M.1 and Laguna XS.2 as free in preview through its OpenAI-compatible API and OpenRouter, though the preview window is described as temporary on the models page.
What models does Poolside offer?
Poolside offers Laguna M.1, a 225B-parameter MoE model with 23B activated parameters, and Laguna XS.2, a 33B-parameter MoE model with 3B activated parameters. Laguna XS.2 is open-weight under Apache 2.0 with weights on Hugging Face.
Where can Poolside be deployed?
Poolside deploys on-premises, inside a cloud VPC, on dedicated enterprise hardware, on a local workstation tower, or on customer-provided NVIDIA GPU clusters via Helm. Air-gapped and disconnected environments are supported for government and defense use cases.
Which IDEs and editors does Poolside support?
Poolside integrates with VS Code, Visual Studio, Zed, and JetBrains through the Agent Client Protocol. It also offers a terminal-native CLI called pool and a browser-based web interface.
Is Poolside API compatible with OpenAI?
Yes. Poolside exposes an OpenAI Chat-compatible API for Laguna models, so teams can point existing OpenAI SDK clients at Poolside endpoints without rewriting integration code.
How do I get access to Laguna M.1 weights?
Laguna XS.2 weights are downloadable from Hugging Face under Apache 2.0. For Laguna M.1, Poolside asks researchers working on behalf of an institution or university to email [email protected] to request weight access or higher API rate limits.