AI assistant

Coming soon

An assistant that works through the same typed actions you do: read-only by default, and every change it proposes is a plan you review and approve.

Built foundations

The groundwork it will stand on is built: every action has typed parameters and a risk level, most changes can preview themselves as a plan, and risky ones ask for a fresh confirmation.

Ask in a side panel or from the command palette. Its tools are the actions you are allowed to run, nothing more.
Changes go plan, diff, your approval (a passkey for high-risk ones), then run, verify and roll back on failure.
Log explainer, config doctor, deliverability fixer, security advisor and migration helper.
Answers grounded in the documentation and this server's own configuration, with secrets removed first.
Your choice of provider (Anthropic, OpenAI, DeepSeek, Mistral, Gemini or a local model) with your own key, or through RespireSoft.