Docs · Start
Quickstart
First chat in under two minutes.
1 · Chat
sentinelType a question, Sentinel streams the answer. Tool calls run in-process and stream back into context.
2 · One-shot from scripts and CI
sentinel ask "why is this test failing?"
sentinel ask -b "fix the typo in src/index.js" # BUILD mode: may edit files3 · Local models
ollama serve
sentinel # auto-discovers installed modelsCost control
Default model is a free-tier model (
openai/gpt-oss-20b via Groq). Every run prints in/out tokens · $ · model; -q silences it.