Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.selftune.dev/llms.txt

Use this file to discover all available pages before exploring further.

Install

Requirements: Bun or Node.js 18+. No extra API keys needed — selftune uses your existing agent subscription for LLM calls.
The examples below use bare selftune commands. If you installed via the CLI only tab, prefix each command with npx selftune@latest (e.g., npx selftune@latest doctor).

Your first session

1. Check health

selftune doctor
This verifies your setup: hooks, config, permissions, and log access. Fix any issues it reports before continuing.

2. Ingest existing sessions

selftune sync
This replays your existing session logs into selftune’s SQLite database. If you’ve been using Claude Code, it will find and process your session history.

3. See your skill health

selftune status
You’ll see a one-line health summary for each skill, plus highlights for skills that need attention.

4. Open the dashboard

selftune dashboard
This opens a local web dashboard with detailed skill health, evolution history, and missed queries.

Guided onboarding

For a step-by-step walkthrough, run:
selftune quickstart
This walks you through init, ingest, status, and suggests next actions based on your skill health.

What happens next

Once installed, selftune works in two modes:

Interactive (agent-driven)

Your agent reads the selftune SKILL.md, routes to the correct workflow, and runs CLI commands. The agent is the operator.

Automated (OS-driven)

System scheduling (cron/launchd/systemd) calls the CLI directly. No agent session needed, no token cost.
Set up automated mode:
selftune cron setup

Next steps

Build Your First Skill

Go from draft skill to measured improvement.

How It Works

Understand the feedback loop architecture.

Core Concepts

Learn about skills, grading, evolution, and evals.

CLI Reference

Full command reference for all 30+ commands.

Cloud Platform

Set up team-wide monitoring with selftune Cloud.