Use Poolside with Ollama

Use Ollama as a local model runtime for Poolside Agent CLI. Ollama can launch pool directly, or you can point pool at Ollama’s OpenAI-compatible API. For a guided setup that installs Ollama, pulls Laguna XS.2, and starts a local agent session, see How to run a Poolside coding agent locally.

Prerequisites

You have Ollama installed.
You have Poolside Agent CLI installed. See Install Poolside Agent CLI.
You have a model available in Ollama.

Steps

Launch through Ollama
Connect directly

Open your project directory:
cd <project-path>
Launch pool through Ollama:
ollama launch pool
This opens Ollama’s model selector. To start with a specific model instead:
ollama launch pool --model <model-name>
To pass flags to pool, add them after --:
ollama launch pool -- --help
Send a prompt. You know the setup works when the agent responds using the model served by Ollama.

Point pool at Ollama’s local OpenAI-compatible endpoint:
POOLSIDE_STANDALONE_BASE_URL=http://localhost:11434/v1 \ POOLSIDE_API_KEY=ollama \ pool --model <model-name>
Use any non-empty value for POOLSIDE_API_KEY when your local Ollama endpoint does not validate API keys.
Send a prompt. You know the setup works when the agent responds using the model served by Ollama.

​Prerequisites

​Steps

​Related resources

Prerequisites

Steps

Related resources