Poolside works where your developers do.
Run agents anywhere your code lives.
DeploymentTerminal-native. Run agents directly from your shell, pipe output into scripts, or automate across repos without leaving the command line.
Where you already work. Agents run inline with your editor; suggestions, edits, and context without switching windows.
Browser-based access. Spin up an agent from anywhere with a full interface. No local setup, nothing to install or configure.
Built for pipelines. Drop agents into CI/CD workflows to automate code review, generation, or testing on every push.
Agents that work within boundaries you set.
Developer featuresSandboxed execution
Every agent runs in an isolated sandbox. Code is executed, tested, and validated in a contained environment before it touches your codebase. No surprises.
Fine-grained permissions
Control exactly what an agent can do. Read a file, write to a directory, run a command, access an API—each permission is explicit and revocable. Your security posture applies to agents the same way it applies to people.
Version control for agent work
Checkpoints layer on top of your existing VCS. Every piece of agent work is a discrete, versioned unit you can inspect, roll back, or roll forward, independent of your Git history. Think of it as save points for AI-assisted work.
A native experience, not a bolt-on
In VS Code, Poolside renders with full theme support. Whether dark or light mode, or your custom theme, it looks like it belongs because it does. Inline Mermaid diagrams render directly in the editor, so architecture decisions and flow charts live next to the code they describe.
Connects to the way you already build.
IntegrationsAgent Client Protocol
Poolside speaks ACP—the open standard for agent-to-editor communication. That means first-class support in any ACP-compatible tool, today and as the ecosystem grows.
❯ pool code
Headless by design, TUI when you want it. pool code fits however you work from the terminal. Run it interactively, automate it, or embed it into your own workflows without touching a GUI.
Developer API
OpenAI-compatible. Swap one line, keep everything else. If you've built on the OpenAI API, you already know how to use Poolside's models—same SDK, same patterns, different endpoint.
Full visibility. Full control.
ConsoleRules and policies
Administrators define the boundaries. Sandbox configurations, MCP server access, and permitted tools are all set centrally and enforced automatically. Developers work freely within guardrails that are already in place.
Agent traceability
Every agent session produces a full trajectory: each action taken, each file touched, each decision made. Visible in the admin dashboard, searchable, and exportable. When compliance asks “What did the AI do?” you have the answer.
Agent observability
Always stay in control of who did what across your organization. Have historical view of all activity in a central pane.
Metrics
Usage data that matters. Track tokens generated, active users, session volume, and cost. All broken down by team, project, or time period.
Models
Use Poolside's foundation models alongside providers you already work with. The Console shows all available models in one view, with Poolside models front and center.
Works with any provider
Poolside Agent isn't locked to Poolside models. Connect any provider or your own self-hosted models—and get the same agent experience.