Skip to main content
Poolside models can help you write, debug, refactor, and automate code in the tools where you already work. Choose a surface, then follow the guide to get started.
Want to try a model before choosing a development workflow? Use chat.poolside.ai to send prompts from your browser. No sign-up required.

Terminal

Run interactive, multi-turn sessions in your project directory.

Automation

Run one-shot tasks in scripts, CI jobs, and other automated workflows.

Editors

Work with Poolside where you write code.

Desktop apps

Use a dedicated app experience outside your editor.

Coding agents

Use Poolside models in third-party coding agents.

Local models

Run Poolside models with a local model runtime.

Applications

Send model requests from your own tools and services.
Some paths run Poolside’s agent directly, such as Poolside Agent CLI and ACP-compatible editors. Other paths use a third-party tool that calls Poolside models through the providers each tool supports.
Third-party tools control their own configuration flows, supported providers, and permission settings. Review the tool’s documentation and security model before you let it read files, edit code, run commands, or use external services.

​
Work from the terminal

Use pool from Poolside Agent CLI for interactive, multi-turn sessions in your project directory. To get started, see Work from the terminal.

​
Automate tasks

Use pool exec from Poolside Agent CLI to run one-shot tasks in scripts, CI jobs, and other automated workflows. To get started, see Automate tasks. To run tasks in GitHub Actions, see Run Poolside in GitHub Actions.

​
Work in an editor

Use Poolside in an editor when you want agent help where you already write code. There are two paths:
  • ACP editors use pool acp to run Poolside Agent CLI from the editor. The editor acts as the ACP client, and pool acp runs the agent session.
  • Editor extensions run their own agent experience and call Poolside models directly or through the providers each tool supports.
Through ACP, editors can expose Poolside Agent CLI features such as plan mode, skills, AGENTS.md instructions, session history, model selection, and slash commands. Editor support varies by ACP client. If your editor does not expose an ACP capability, Poolside may still support the feature, but it may not appear in that editor’s UI. For setup guides, see Editors.

​
Build applications

Use the Poolside API when you want to call Poolside models from your own application, internal tool, script, or service. The API is OpenAI-compatible, so you can call it directly or use the OpenAI SDK with the Poolside base URL and API key. For request examples, streaming, and tool use, see OpenAI-compatible API examples.

​
Next steps

After you choose where to work, see Configure to shape agent behavior and control what it can do. For a guided local setup, see How to run a Poolside model locally with Ollama. To evaluate the experimental vLLM Metal setup on Apple silicon, see How to run a Poolside model locally with vLLM Metal on a Mac. For Ollama integration options, see Use Poolside with Ollama.