WorkTracker › Features › AI summaries

AI summaries at every level — and a daily review

Every work item gets a one-line summary in your language; every day gets a tracking check and suggested follow-ups. Model and provider are recorded on everything.

Summaries where you need them

Sent mail, chat conversations, document edits and grouped commits each get a short AI summary, shown on hover — what it was about without opening five apps. Summaries are written in your preferred language.

The daily review

The tracking check reads the whole day and lists what looks trackable but has no time on it yet. Suggested todos ask the complementary question — what should you do next — and convert to real todos with one click.

Auditability built in

• Every stored summary records which provider and model wrote it.
• Summaries go stale visibly when the underlying events change.
• Contents are fetched transiently for the AI call and never stored — only the summary is.

Your AI, your rules

Each user picks the AI that reads their content: OpenAI, Anthropic, Google Gemini, or any self-hosted OpenAI-compatible endpoint (Ollama and friends) — so analysis can stay entirely on hardware you control. Every call runs behind the layered privacy filter described at /privacy.