PocketAgent × Ollama

Chat with a model running locally on your machine via Ollama. Your installed PocketAgent is automatically prepended to the system prompt. Nothing leaves your device.

Ollama server

host

model not connected

first time? Install Ollama and enable CORS for this origin so the browser can talk to it:

OLLAMA_ORIGINS="https://johnjboren.github.io,http://localhost:*" ollama serve

Then pull at least one model: ollama pull llama3.2.

Active agent

Install an agent from the gallery or paste a #pa=… URL anywhere on this site. The agent is prepended to every request below.

Conversation

← PocketAgent landing · spec · WebLLM chat (zero install)