Quickstart: The Plugin (Claude Code)¶
The fastest way to use Attune AI is as a Claude Code plugin. Install it once, then drive everything — code review, security audits, test generation, spec-driven development — with plain English in your normal Claude Code session.
No CLI to learn. No Python package. No separate API key — the plugin runs inside the Claude Code session you already have.
Prefer the CLI or Python?
This page is the plugin-first path. If you'd rather run workflows from a
terminal or from code, start with First Steps instead —
it covers the attune CLI and the Python API.
What you get¶
The plugin ships 17 auto-triggering skills. You don't memorize commands — you describe what you want, and the right skill activates.
| Say something like… | Skill that activates | What it does |
|---|---|---|
| "review this file for quality issues" | code-quality |
Code review + bug prediction |
| "scan src/ for vulnerabilities" | security-audit |
eval/exec, secrets, injection, path traversal |
| "what tests am I missing?" | smart-test |
Finds coverage gaps, generates tests |
| "this test keeps failing" | fix-test |
Diagnoses and fixes, up to 3 attempts |
| "help me plan this feature" | planning |
Architecture + TDD planning |
| "let's build X from a spec" | spec |
Brainstorm → plan → execute with quality gates |
| "what can attune do?" | attune-hub |
Routes you to the right skill |
Step 1 — Install the plugin (1 min)¶
In Claude Code:
That's it. The skills are now available in every Claude Code session — no API key, no config file.
Step 2 — Orient yourself (30 sec)¶
Ask the hub what's available:
The attune-hub skill activates and routes you to the workflow that fits your
goal. Use this whenever you're not sure which capability you want.
Step 3 — Run your first real workflow (3 min)¶
Point it at code you actually care about. Two good first runs:
Code review — open a file (or just name it) and ask:
The code-quality skill spins up a small team of specialist subagents, reads
your real source, and returns prioritized findings with file/line references
and suggested fixes.
Security audit — for a vulnerability-focused pass:
The security-audit skill scans for eval/exec misuse, hardcoded secrets,
path traversal, and injection risks, then reports each finding with severity
and a fix.
Either one shows you the core pattern in a couple of minutes: describe the goal → specialist agents review your real code → you get actionable, cited findings.
Step 4 — Go deeper (optional)¶
Once the basics click, the highest-leverage skill is spec-driven development:
This walks you from a rough idea through requirements, design, and an ordered task list — with human approval gates between phases — then executes the tasks. It's the recommended path for anything non-trivial.
Other skills worth trying by name or description:
smart-test— "find untested code in this module and write tests for it"refactor-plan— "this file has too much going on, plan a refactor"release-prep— "get me ready to cut a release"recall— "did I hit this problem in a past session?"
What's next¶
| If you want to… | Go to |
|---|---|
| Run workflows from a terminal or Python | First Steps |
| Add the full MCP toolset (41 tools) + CLI | MCP Integration |
| Pick a longer learning path | Choose Your Path |
| See every workflow | First Steps → Try More Workflows |
Plugin vs. package
The plugin gives you the 17 natural-language skills with zero setup.
Installing the Python package (pip install attune-ai) adds the
attune CLI, the MCP server, and 41 MCP tools on top. You can start with
the plugin today and add the package later — they layer cleanly.