Getting started

Install

Alden is in open beta: anyone can install it, and it's free during the beta. Pick one of these.

The desktop app, on macOS. Alden in a window and the menu bar, which briefs your top PRs while you're away. It's signed and notarized, runs on Apple silicon and Intel (macOS 12 or later), and installs the alden command too, so it can be your only install. See The desktop app.

brew install --cask getalden/tap/alden-desktop

Or download Alden-<version>-universal.dmg from the releases page.

Homebrew, on macOS or Linux. The alden command as a standalone binary, which needs nothing else.

brew install getalden/tap/alden

npm, anywhere with Node 22.13 or later. During the beta, releases are published under the next tag.

npm install -g alden@next
npx alden@next --help                # or run it without installing

A standalone binary. For macOS (Apple silicon and Intel), Linux (x64 and arm64) and Windows (x64), from the releases page: alden-<version>-<platform>.tar.gz, or .zip for Windows, with a SHA256SUMS to check it against. Unpack it and put alden (alden.exe) on your PATH.

The examples below assume alden is installed; with npx, put npx alden@next in place of alden.

Start

alden

On its own, signed out, alden walks you through the same three steps as the desktop app's and the web UI's first run, in text: it signs you in to GitHub, asks how Alden should think (an Anthropic or OpenAI key, a local model, or checks only; you can say no), then opens your queue. After that, alden opens the interactive view: your queue, briefings, feedback, comments and sending, on one screen. See alden (interactive).

The desktop app and alden ui start with the same three steps when there's no sign-in yet.

To sign in without the interactive view:

alden auth login

If the GitHub CLI (gh) is signed in, Alden uses its token and stores nothing. Otherwise Alden shows a code to enter at github.com, and stores the resulting token in your OS keychain. alden auth status shows which account and token Alden is using. See Commands for the other options.

Alden asks for the repo scope (to read private repos) and read:org (to see review requests sent to your teams).

At a company? Alden needs no install on your organisation: it reads what your own sign-in can read. But many organisations use single sign-on or only allow approved apps, and then hide their PRs from a sign-in until it's authorized. Run alden doctor after signing in: it names any organisation whose PRs Alden can't see, and how to fix it. If your company already uses the GitHub CLI, its sign-in is usually the quickest route. See Troubleshooting.

See what's waiting on you

alden queue

The PRs waiting on your review, in two groups: Needs you, ranked, and Fast-track, for small low-risk PRs you can clear quickly. Every row says why it ranks where it does. Open one with alden open <#>. See The review queue.

Above them, Your turn lists the PRs you already reviewed where something happened since: a reply to your comment, a thread resolved, a push that outdated your comment. Opening one takes it off the list. See Your turn.

PRs written by AI coding agents (Copilot, Claude Code, Cursor, Devin, Codex and others) say so on their row, including a person's PR that an agent wrote, and how Alden knew (its branch, its description or its last commit). alden queue --author agents shows only those. See Bots and AI agents.

Get a briefing

alden review 1                       # #1 from your last queue
alden review owner/repo#123          # or any PR
alden review https://github.com/owner/repo/pull/123
alden review                         # or your own uncommitted changes

A briefing has three parts:

Needs your eyes: a few places (3 unless you change review.eyes), each with a file, a line and the reason.
Checked: every check Alden ran, what it found, and the evidence.
Not checked: what Alden couldn't check, so you know what's still on you.

See Briefings and checks.

With a model, you can also ask about a PR once it's briefed: alden ask 1 "Is the retry safe?" answers from the briefing, the diff around the places that matter and your notes, in one call within your caps. See Ask about this PR.

Or in your browser

alden ui

The queue, briefings, your feedback and what Alden learned, in a local web page. A PR reads in three steps, Brief, Walk through (the flagged places one at a time) and Diff; the Comments drawer holds your drafts until you send them as one review, and Tune picks the model, the tone of Alden's drafts and asks the model a question. See Alden in your browser.

Add a model (optional)

Without a model, Alden runs only its deterministic checks. With one, the briefing also says what the change does, whether it matches its title, and what else deserves a look.

alden setup

It offers three choices, and takes about a minute:

Paste an API key, Anthropic's (create one at console.anthropic.com/settings/keys) or OpenAI's (at platform.openai.com/api-keys): Alden tells them apart. It checks the key with one tiny request, which costs a fraction of a cent, and keeps it in your keychain, where the desktop app finds it too.
Use a local model with Ollama. Alden suggests a coder model that fits your machine's memory, downloads it if Ollama doesn't have it, and points its config at it. Free and private, but local models flag less than Claude.
Checks only. Nothing to set up.

In the web UI and the desktop app, Set up a model (where the briefing options would ask the model) offers the same. alden auth model stores a key without the check, and ANTHROPIC_API_KEY or OPENAI_API_KEY works too, coming first where it's set. If you'd rather Alden handled everything, sign up for Alden Cloud (beta).

With a model, Alden can also draft comments on the places it flagged, in a tone you pick: terse, concise, detailed or teaching (alden config set drafts.tone detailed, or Tune in the UI per repo). Its drafts wait for you to keep, edit or drop them, and none is sent untouched. See The tone of drafts.

The default is Claude Sonnet, capped at $0.50 per review and $10 per day. Each briefing shows what it cost, and alden spend shows the running total. See Models, providers and budgets for other models, AWS Bedrock, Google Vertex, Azure Foundry and local models such as Ollama.

Get callers outside the diff

For TypeScript, JavaScript, Python, Go and PHP, Alden builds a code graph of the repo and lists who calls the code a change touches. For local changes this just works. For a PR, run alden review from inside a clone of that PR's repo:

cd ~/code/their-repo
alden review 123

Alden remembers each clone you run it in, so later briefings of that repo's PRs find callers from anywhere. If your clones live in one folder, alden config set clones.folders ~/src lets Alden find them all. See The code graph.

In Claude Code

With Alden installed and signed in, add the plugin in Claude Code:

/plugin marketplace add getalden/claude-plugin
/plugin install alden@getalden

(Or from a terminal: claude plugin marketplace add getalden/claude-plugin, then claude plugin install alden@getalden.) Then /alden:review and /alden:queue. See Claude Code plugin.

Get help

Ask in the Alden Community on Discord, where you can also tell us what works and what doesn't. For errors, start with alden doctor and Troubleshooting.