Poolside — Foundation models for long-horizon software development agents. Best for Enterprises building mission-critical software in regulated industries like defense and finance, Teams needing AI that can plan and execute multi-step development tasks over hours or days, Organizations requiring on-premise or air-gapped AI deployment with no data leaving control. Contact Sales pricing.
Enterprises building mission-critical software in regulated industries like defense and financeTeams needing AI that can plan and execute multi-step development tasks over hours or daysOrganizations requiring on-premise or air-gapped AI deployment with no data leaving controlCISO-led adoptions where model governance, tracing, and role-based access are mandatoryForward-thinking enterprises willing to co-develop AI outcomes with embedded research engineers
Not ideal for
Individual developers or small teams seeking a self-serve code completion toolProjects requiring ChatGPT-like conversational AI rather than software-focused agentsTeams unwilling to invest time in co-developing custom evaluations and workflowsOrganizations with simple coding needs where generic copilots sufficeBuyers expecting transparent, per-seat pricing or a free trial tier
Poolside is a compelling option for security-conscious enterprises needing AI for complex software workflows. Its on-prem deployment and joint ownership model reduce risk, but the high-touch approach may not suit teams seeking a plug-and-play copilot.
Poolside stands out by targeting the hardest problems: long-horizon, multi-step software development in high-consequence environments. Its foundation models are trained with reinforcement learning to reason like humans, not just autocomplete. For enterprises in defense, finance, or critical infrastructure, Poolside's ability to deploy inside your VPC or air-gapped network is a decisive advantage. The forward-deployed research engineer model ensures adoption but limits scalability—this is not a self-serve tool. Compared to GitHub Copilot or Codeium, Poolside offers deeper reasoning but higher cost and complexity. Pass if you need a quick code assistant; pick if you're building mission-critical systems where failure is not an option. The lack of transparent pricing and limited public integrations may slow evaluation. Real-world caveat: success depends heavily on the quality of your internal data and the patience to co-develop evaluations.
Skip Poolside if Skip Poolside if you need a self-serve, low-cost code completion tool for individual use or small teams without on-premise infrastructure.
Latest from Poolside
Updated yesterday
Across the latest 7 updates: 2 feature updates, 3 launches, 1 community discussion and 1 news mention.
How likely is Poolside to still be operational in 12 months? Based on 6 signals including funding, development activity, and platform risk.
funding runway
60
website health
90
github activity
50
category mortality
70
wrapper dependency
100
hyperscaler overlap
80
About Poolside
Poolside is a frontier AI lab building foundation models purpose-built for software development and enterprise-grade agent systems. It targets organizations that need AI to handle complex, multi-step reasoning tasks across long time horizons—starting with software engineering. Unlike generic coding assistants, Poolside combines single and multi-agent orchestration with policy governance, sandboxed execution, and end-to-end tracing. Its forward-deployed research engineers embed with customer teams to co-design and operate models inside the enterprise's own security boundary, ensuring outcomes over token counts. Key features include fine-tuned foundation models deployable on-prem or in VPC, connectors to repositories and data warehouses, and executive-grade risk controls aligned with CISO requirements. Poolside positions itself as a secure, outcome-driven alternative for enterprises that cannot accept the data leak risks or shallow code completion of consumer AI tools.
Researching Poolside? Skip the guesswork.
Tell us what you want to build — we'll match the AI tools that fit your goal, budget & existing stack.
Key Features
Foundation models fine-tuned for software development
Single and multi-agent orchestration
Sandboxed execution environments for agents
Governance policies with end-to-end traces
Role-based access control for humans and agents
Connectors to repositories, databases, data warehouses
On-prem, VPC, or air-gapped deployment
Forward-deployed research engineers embed with teams
Joint responsibility for measurable business outcomes
Executive-grade risk controls and auditability
IDE extensions, TUI, binaries, and workflows
Real-world workflow fit
Concrete scenarios for the personas Poolside actually fits — and what changes day-one when you adopt it.
Security-conscious enterprise dev team lead
Your team needs to automate code reviews across a multi-cloud environment without exposing proprietary code to external APIs.
Outcome: Poolside deploys Laguna models in your VPC with connectors to your private repos. Multi-agent system automatically flags vulnerabilities and suggests fixes, while role-based access control ensures only authorized humans and agents can approve changes. Full audit trail satisfies CISO.
Federal contractor with air-gapped systems
You need to modernize legacy software on an air-gapped network with no internet access. No commercial AI tool can be used.
Outcome: Poolside installs its runtime on-site (defense workstations). Forward Deployed Research Engineers embed with your team to build custom agents for code migration and testing. Models run entirely inside your boundary, with no data egress. Outcome ownership ensures measurable modernization.
Enterprise architect planning a large-scale system migration
You're migrating from a monolithic Java app to microservices across multiple clouds. The project involves hundreds of developers and years of technical debt.
Outcome: Poolside's multi-agent orchestration (Fern Labs Bridge) coordinates dozens of agents: some plan the migration steps, others refactor code, and a separate agent runs tests. Agents are governed by policies and end-to-end traces. The migration timeline is compressed via parallel agent work.
Maintain full audit trails and governance over AI-assisted code changes in regulated industries.
Utilize multi-agent orchestration (Fern Labs Bridge) to automate end-to-end testing and deployment pipelines.
Run evaluation benchmarks collaboratively with Poolside to tailor models to your specific codebase.
Models Under the Hood
Laguna M.1Laguna XS.2
Limitations
Poolside's models and products are currently in research preview, so availability and feature completeness are limited. The platform requires close collaboration with Poolside's Forward Deployed Research Engineers, which may not suit teams wanting a self-serve tool. No public pricing is available, and on-premise deployment may entail significant infrastructure requirements. The focus on enterprise and government customers means individual developers and small teams are not the target audience.
Hidden costs & gotchas
What the public pricing page doesn't put in bold. Captured from pricing-page footnotes, contract terms, and recurring complaints.
•Opaque pricing: all engagements are contact-sales, likely six-figure+ annual contracts.
•Infrastructure costs: you provide and maintain on-premise or VPC compute (GPUs, storage, networking).
•Engineering time: setup and customization require weeks of collaboration with Poolside engineers.
•No free tier or trial: you cannot evaluate Poolside without a paid engagement.
Where the pricing makes sense
The company stage and team size where Poolside's pricing actually pencils out — and where peers do it cheaper.
Poolside's pricing is contact-only, targeting large enterprises and government agencies. There is no public tier, so it's likely the most expensive option in the AI coding space. For comparison, GitHub Copilot costs $10-39/user/month, Cursor is $20/user/month, and Amazon CodeWhisperer is free for individuals. Only choose Poolside if you need on-premise security, custom models, and outcome-based partnerships where the cost is justified by data sovereignty and governance requirements.
Setup time & first value
How long it actually takes to get something useful out of Poolside — broken out by persona, not the marketing-page minute.
Setup is not self-serve. Expect a multi-week to multi-month engagement: initial discovery and environment assessment (1-2 weeks), model deployment and connector setup (2-4 weeks), co-evaluation and custom agent development (4-8 weeks), then ongoing optimization. Poolside's Forward Deployed Research Engineers embed full-time during later phases.
Switching to or from Poolside
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
Migrating in
→From GitHub Copilot: Poolside's engineers will help transition from cloud-based code completion to on-premise agents, including setting up custom evaluation benchmarks and agent workflows.
→From manual code review processes: Poolside's agent platform can automate review pipelines; engineers co-configure policies and audit trails to replace manual checks.
→From in-house custom agents: Poolside can ingest your existing agent scripts and run them on its runtime, adding governance and multi-agent orchestration.
Migrating out
↗To GitHub Copilot: Export agent traces and custom evaluations; Copilot does not support on-premise models, so data sovereignty is lost.
↗To open-source agent frameworks (e.g., LangChain): Poolside's proprietary models and runtime cannot be directly exported; you would need to rebuild workflows using open models.
↗To a different AI vendor: Poolside's embedded engineers can assist in knowledge transfer of custom agent logic, but the models and traces remain proprietary.
Recent material changes
Pricing, brand, ownership, or deprecation changes worth knowing before you commit. Most-recent first.
•2026-05-11: Published 'Through the looking glass of benchmark hacking' blog outlining reward hacking challenges in training evaluations.
•2026-05-05: Announced the Poolside Platform — production-grade AI agents inside your security boundary with full auditability and governance.
•2026-04-28: Released Laguna M.1 and XS.2 models, along with a runtime for training and operating agents, in research preview.
•2025-11-18: Acquired Fern Labs (London-based forward-deployed research engineering company) to enhance multi-agent orchestration with Bridge technology.
