Skip to content
Poolside

World-class AI, every layer yours.

Poolside delivers a complete stack of world-class models, agents, APIs, and applications—deployable into the infrastructure you control, with customisable pipelines, enterprise-grade governance, and no third-party dependencies. Every layer sovereign, entirely yours.

Available on AWS, Azure, Google, and on-prem Talk to an Expert

Don't rent intelligence. Own it.

Deploy anywhere

Poolside was built security-first. Air-gapped and on-premises deployment aren't afterthoughts, they're the foundation. You receive full model weights, not just API access, making your AI capability a permanent sovereign asset.

Whether you're a nation securing strategic infrastructure or an enterprise protecting competitive advantage, every query and every workflow runs entirely within your infrastructure. No external calls. No data egress. No shared systems.

Full stack

The complete platform—foundation models, Console, developer tools, and infrastructure—deployed as an integrated system inside your boundary. Everything runs behind your perimeter, managed by your team.

Cloud VPC

Deploy Poolside into your existing cloud environment. Self-managed, within your account, under your policies. No data leaves your VPC.

On-Prem Enterprise rack

Pre-provisioned hardware for teams that need scale without external dependencies. A single-node deployment supporting hundreds of developers, fully air-gap capable.

On-Prem Tower

For small classified teams or constrained environments. A compact, quiet workstation that runs Poolside locally with full offline operation.

Bring your own hardware

Already have GPU infrastructure? Poolside installs on customer-provided clusters powered by NVIDIA GPUs. Your hardware, our software—deployed via Helm, aligned with enterprise-standard practices.
  • Air-Gapped and Disconnected
  • STIG-Hardened OS
  • ATO Achieved
  • Cleared Resources
  • IL5 Deployable
  • On-Prem/GovCloud/Edge

Every action. Traced, governed, exportable.

Auditability and control

AI agents that touch production systems need more than permissions, they need proof. The Poolside Console gives CISOs, CTOs and General Counsel a single pane of glass for everything agents do across the organisation—with the audit trails, governance controls, and exportable records each of them needs.

Traceability

Every agent session produces a complete trajectory—each tool call, file edit, reasoning step, and decision recorded and searchable. When an auditor asks what the AI did, you hand them the log.
  • Full action-by-action timeline per session
  • Expandable detail on every step
  • Exportable for compliance workflows

Sandboxed execution

Agents run in isolated containers with explicit boundaries. Filesystem access, network policies, and credential injection are all defined before the agent starts—not negotiated at runtime.
  • Custom container images supported
  • Private container registries
  • Network access controls per sandbox

Secrets and redaction

Credentials are centralised, encrypted at rest, and referenced by ID. They're injected at runtime and automatically redacted from agent outputs—by exact match and by pattern.

Secrets never appear in logs, trajectories, or model context.

Audit retention and SIEM integration

All API operations tracked in accordance with NIST best practices. Configurable retention periods, cold storage for historical logs, and direct integration with your existing SIEM. Event propagation in under five minutes.

Fine-grained permissions and Role-Based Access Control RBAC

Control what each agent can do at a granular level—per agent, per team. Read-only, read-write, shell access, MCP tool access.

The same rigor you apply to employee access, applied to AI.

Administrator-defined rules for agents and developers

Sandbox configurations, permitted MCP servers, available tools, and agent step limits—all defined centrally by administrators and enforced automatically. Developers work freely within guardrails that are already in place.

Who we work with.

Outcomes, not tokens. Inside your boundary.

Custom solutions
two engineers working side by side at a whiteboard—one Poolside, one customer—sketching out a workflow architecture

We don't hand off models. Forward Deployed Research Engineers embed with your product, engineering, and platform teams to take joint responsibility for outcomes, adoption, and measurable, long-lasting business impact.

Don't rent intelligence. For critical systems where a generic integration won't cut it, we're the only solution built to take joint ownership of the outcome.

two engineers working side by side at a whiteboard—one Poolside, one customer—sketching out a workflow architecture

How it works.

  • FDREs are AI-native research engineers who sit with your team; not in a support queue, not on a ticket. They take joint responsibility for outcomes, adoption and measurable, long-lasting business impact.

    They work across multi-cloud, legacy, and air-gapped environments. Ensurig risk controls and auditability are co-designed with you in the toughest environments, not bolted on after the fact.

  • Use-case focused process improvement and pain-point elimination. Walk through your workflows, identify where agents create the most value, and define success metrics with stakeholders from day one.
    Pain points identified. Clear scope, defined targets, stakeholder alignment.
  • Engineers integrate with your product, engineering, and platform teams. They learn your domain, understand your data, and build on Poolside's platform primitives—sandboxed execution, centralized credential management, full auditability, and code-first orchestration.
    Team integrated. Your team and ours working as one.
  • Deliver agents, APIs, and developer surfaces in a matter of weeks. The first automated workflow goes live—running in your environment, under your controls.
    First workflow live. Production running daily.
  • Hardening, observability, and adoption programmes ensure reliability and measurable ROI. This isn't a handoff—it's a go/no-go decision backed by evidence, with a path to expand.
    ROI measured. Expansion plan, or an honest no-go.

We don't hand off models

We take joint responsibility for outcomes—from assessment through production—using the same platform your team will operate going forward. Every workflow we build is code you own, running on infrastructure you control. Your data never leaves your boundary.

Public Sector

Deploy sovereign AI that never leaves your infrastructure. Air-gapped, compliant, and cleared for sensitive workloads across civilian agencies. → Poolside in the Public Sector

Most processes only exist in someone's head.

Custom agentic pipelines

Built around your systems

Connect to the tools, APIs, and data sources your teams already use, across engineering, operations, go-to-market, and beyond. No rip-and-replace. Workflows are modeled around how your organization actually works, not how a generic template assumes it does.

Orchestrated, not improvised

Every pipeline is code-first and fully auditable. The institutional knowledge that used to live in a senior engineer's instincts; the edge cases, the exceptions, the unwritten rules, gets encoded into defined logic. Consistent outputs, regardless of who's in the loop.

Self-improving by design

Organizations change. Pipelines should too. Meta-pipelines continuously review agent trajectories, surface what's working and what isn't, and update core workflows to reflect new standards and accumulated learnings. The longer they run, the sharper they get, and everything they learn stays yours.

See how Poolside deploys in your environment.

Talk to us about your infrastructure, compliance needs, and where agents can have the most impact.
Talk to an Expert
Also available on AWS, Azure, Google, and on-prem