Power Claude quick setup runs backwards on purpose. Most onboarding wizards angle to sell you the biggest plan on the way in; this one — a 90-second "Consult" — tries to talk you into spending less. Add one account, answer a single goal question, read the Seat Ledger, and decide with your real numbers before you commit a dollar.
Ninety seconds, and the wizard argues you down
Here is the pitch nobody makes out loud: spend less than you probably are. If you're either eating rate-limit stalls on one over-worked account, or paying for a Claude Max-20x ceiling you never actually sustain, the Consult exists to point that out — with your numbers, before you commit a dollar.
The Consult is the getting-started screen that runs the first time Power Claude activates. It's a short, opinionated interview: it asks one question about what you're optimizing for, then translates your answer into a concrete plan — how many Claude accounts ("seats") to pool, and in what mode. That's the whole ceremony. It is deliberately not a firehose of feature toggles. Everything advanced — Balanced/Power Mode, the watchdog, pre-emptive rotation, session snapshots — is already wired sensibly by default. The Consult's only job is to right-size the one decision that actually costs money.
This is a mechanism, not a marketing funnel. No dark patterns, no "add five accounts now for maximum savings!" nudge on the way in. And one thing up front, because setup is exactly the moment you should be suspicious: Power Claude is an independent tool built for Claude Code. It is not affiliated with or endorsed by Anthropic.
If you'd rather have the slow, click-by-click, every-gotcha version, that lives in a companion piece linked at the bottom. This article is the fast, decision-first pass.
The 90-second flow
Cold install to a working pool, in order:
- Install. Grab the extension from the downloads page or the VS Code Marketplace. First activation creates
~/.power-claude/and starts a local rotation proxy — a zero-dependency Node proxy, nothing tonpm install. - Activate. Paste your license key, or let the built-in trial start. The Consult opens itself once the extension is live.
- Add one account. Click Add Account (or run
pc add), then sign into a Claude.ai account you own through the normal OAuth flow. One is enough to finish onboarding — you can add more later. - Answer the one goal question. This is the only decision the Consult forces on you.
- Read the Seat Ledger. It shows the money math for your goal against the accounts you added — a floor, not a ceiling.
- Start lean. Accept a minimal pool now; your real usage tells you when to add a seat.
Prefer a terminal to the panel? The CLI mirrors the whole thing:
Command Palette → "Power Claude: Open Getting Started"
...or, from any terminal:
pc onboard # (re-)open the Consult
pc add # add a Claude account
Six steps, one real decision. The rest is just clicking through.
The one question that changes the bill
The Consult asks what you're optimizing for. There are three honest doors, and each does exactly what it says on the label — no mystery box behind any of them:
- Maximize usage. Highest sustained throughput, and you don't mind paying for it. Leans toward Balanced/Power Mode across more seats, because effective throughput scales near-linearly with account count — each account carries its own independent 5-hour, weekly, and per-minute windows.
- Save money. You suspect you're overpaying. This is the right-sizing path, and it genuinely produces a smaller recommendation than the other two.
- Balanced. A sane middle: enough seats to stop the stalls, not so many that you're renting headroom you never touch.
There is no wrong answer here, and none of them is a trap that quietly upsells you. "Save money" really does recommend fewer seats than "Maximize usage." That contradiction — a setup wizard that talks itself down — is the entire reason the question exists. Think of it as the only honest game show on television: every door has its prize printed on the front in plain text.
The Seat Ledger: see the math before you spend
The Seat Ledger is where the Consult earns its keep. It lays out, line by line, what pooling a handful of Claude Pro seats does to your bill versus the plan you're on now — the itemized receipt, with the line you never actually used crossed out in red pen.
The headline the tool uses as its source of truth: pooling roughly five Claude Pro accounts instead of paying for a single Claude Max-20x plan puts about $100+/mo back in your pocket — and that figure is a floor, not a ceiling. Because pooling has no fixed cap, the savings grow with every Max plan or extra account you retire. Pro billed annually is about 17% cheaper, which pushes the five-account basis further past $100.
Read the justification carefully, because it is not a raw-limit claim. Anthropic names the top plan 20x for a reason, and this is not the assertion that five Pro accounts equal that raw ceiling. It's that most Max-20x subscribers never actually sustain their ceiling — you're paying for headroom you don't burn. Five independent Pro windows cover what you actually use. The Ledger right-sizes your spend to your real effective usage, not to a number on a marketing page.
And the caveat the Ledger carries in bold, because honesty has to cut both ways: if you genuinely saturate a Max ceiling every single day, staying on Max may be the right call. Right-sizing is for the majority who don't. The Consult would rather leave you on the plan that fits than churn you into a worse one just to prove a point.
One line item that never gets folded into the savings headline: Power Claude itself is about $10/mo. That's a tool cost, not a discount on your Anthropic bill, and the Ledger shows it on its own row so the math stays honest. (Turn on the token-saver and it trims roughly 25%+ off the bulky tool-output it compresses — that's fewer tokens on the wire, not dollars off your bill.)
The Seat Ledger recomputes every time you add an account, so each new seat's payoff is visible before you commit to it. For the full, spreadsheet-grade version — every assumption spelled out — the deep dive lives at Claude Max vs. pooling Pro accounts: the real cost.
Start lean — add a seat only when usage proves it
Here's the ramp-up the Consult is built around, and it's the least salesy thing about the whole tool: don't front-load five accounts on day one. Add one, run your normal week, and let the dashboard tell you the truth. If you never brush your limits, you never needed the second seat. If you're stalling out by Wednesday, the runout forecasting makes that obvious — and then you add a seat.
This inverts the usual SaaS incentive. You add capacity when your real usage demonstrates the need, not because a wizard told you "more is better" before you'd typed a single prompt. Reading those signals — which account is warm, when a runout is coming, how the pool is balancing — is covered in how the rotation dashboard actually works.
What actually leaves your machine
Setup is the right moment to be suspicious, so here's the plain disclosure with no hand-waving:
- The proxy is local. It routes requests from your machine directly to Anthropic's API, the same way Claude Code already does. It is not a relay to another cloud — your prompts and completions do not pass through a Neural-LLM server.
- You own every account you add. Power Claude does not supply, share, or pool accounts with strangers. Each account is used strictly within its own published limits, and rotation is pre-emptive — it moves you to a fresh seat before a 429, reading the rate-limit headers, rather than crashing into the wall and recovering afterward.
- Credentials stay encrypted, locally. Your Claude account credentials and tokens are encrypted on your machine and never leave it — they're used only to talk to Anthropic. The one outbound call Power Claude makes is licensing: a lightweight, signed check to the license server (with a device fingerprint so a license stays bound to the machine that activated it), and it keeps working offline for a grace period if that server is ever unreachable. No prompts, no files, no transcripts.
That's the whole surface. Nothing about your work leaves the box.
Want the full, click-by-click walkthrough?
This article is the fast pass — the decisions, the money math, and the 90 seconds it takes to get a right-sized pool running. If you'd rather have every step spelled out, with the prerequisites, the reload prompt everyone skips, the credential-file permission trap, and how to watch your first rotation happen without the editor pausing, read the companion piece: Setting Up Power Claude in VS Code — a 10-minute walkthrough. The two are complementary: start here to decide how many seats, go there to nail every install detail.
And the Consult is never a one-shot. Re-open the whole flow any time from the Command Palette ("Power Claude: Open Getting Started") or with pc onboard in a terminal. It's idempotent — it re-reads your current accounts and re-renders the Seat Ledger with today's numbers, which makes it genuinely worth re-running after your usage shifts or after you retire a Max plan.
FAQ
What is Power Claude quick setup?
It's the "Consult" — a 90-second getting-started flow that runs the first time Power Claude activates. Instead of dumping feature toggles on you, it asks one question about what you're optimizing for (maximize usage / save money / balanced), then shows a Seat Ledger with the money math for your goal against the accounts you added. Advanced features are already on by default, so onboarding only asks the one decision that actually costs money: how many Claude accounts to pool.
How long does Power Claude setup take?
About 90 seconds for the first pass: install, activate (or start the trial), add one account via OAuth, answer the single goal question, and read the Seat Ledger. That gets you a working, right-sized pool. Adding more accounts and fine-tuning modes is optional and happens later from the dashboard — none of it blocks you from getting started.
How many Claude accounts do I need to start with Power Claude?
One is enough to finish onboarding and see the tool work. The Consult deliberately steers you to start lean and add a second seat only when your real usage proves you need it — the dashboard's runout forecasting makes that obvious. Front-loading five accounts on day one is exactly the thing this onboarding is designed to talk you out of.
What is the Seat Ledger?
The Seat Ledger is the itemized cost view the Consult shows you: what each pooled seat costs and what it buys, versus the plan you're on now. It recomputes every time you add an account, so a new seat's payoff is visible before you commit. Its source-of-truth headline is that pooling roughly five Claude Pro accounts instead of one Max-20x plan puts about $100+/mo back in your pocket — a floor that grows as you retire more plans. Power Claude's own ~$10/mo shows on its own separate row so nothing is hidden.
Does Power Claude actually save money versus Claude Max?
For most people, yes — as right-sizing, not raw-limit equivalence. Pooling roughly five Claude Pro accounts instead of one Claude Max-20x plan puts about $100+/mo back in your pocket, a floor that grows as you retire more plans (Pro annual is ~17% cheaper, pushing it further). It works because most Max-20x subscribers never sustain their ceiling — five independent Pro windows cover what you actually use. The honest caveat: if you genuinely saturate a Max ceiling every single day, staying on Max may be the right call. Power Claude's own ~$10/mo is a separate line item, never folded into the savings headline. Full math lives at Claude Max vs. pooling Pro accounts.
Does Power Claude see my prompts or store my credentials?
No. The rotation proxy is local — your prompts and completions go straight from your machine to Anthropic's API, not through any Neural-LLM server. Your Claude credentials and tokens are encrypted on your machine and never leave it; they're used only to talk to Anthropic. The single outbound call Power Claude makes is a signed license check — no prompts, files, or transcripts included. Power Claude is an independent tool and is not affiliated with or endorsed by Anthropic.