Skip to content
SGPT

Introduction

SGPT is a command-line AI coding agent. You open a chat in your terminal, give it a task, and it reads files, runs commands, edits code and searches your team’s knowledge base — pausing for your approval before anything with side effects.

What makes it different

Section titled “What makes it different”

It is configured by the repo, not the user. Roles, tool sets and lores live in .sgpt/ directories committed with the code. Clone a repo and you get its personas, its integrations and everything the team has learned.

It talks to one backend. SGPT is a thin client: chats, messages and model routing live behind a gRPC AiService. The CLI mirrors messages for display and streams generations; the server persists and bills.

Tools are protobuf services. The built-in tools are declared as a gRPC service whose descriptor is embedded in the binary — descriptions come from proto comments, JSON schemas from request messages, and auto-execution from the idempotency_level. Remote tool engines expose any other service the same way.

Knowledge is a first-class artifact. Lores are markdown files with YAML front matter, searched with a regex, and written only with your approval.

Where to go next

Section titled “Where to go next”