Install and sign in
Run one command — npx fetches the CLI, so there's nothing to install first. A browser window signs you in to your free account.
Works with Claude Code, Codex & Cursor · Updated June 2026
Run Diffmode inside Claude Code, Codex, or your terminal.
The same hosted pipeline the web app runs, driven by your AI agent or run by hand. Start with a free audit — made for bootstrapped teams.
npx diffmode@latest loginFree account, no credit card. Your first audit is covered by the credit you get at signup.
From sign-in to a growth plan, in four moves
Same pipeline as the web app — you just drive it from your terminal or your AI agent instead of the browser.
Add the skill
One more command teaches Claude Code, Codex, and Cursor how to drive the CLI for you — sign-in, credits, the run, and the results.
Ask your agent, or run a command
Tell your agent to run a growth plan for your URL, or run the command yourself. The first audit is covered by the credit you get at signup.
Read it and apply it
You get your landscape, advantages, buyers, blockers, and a starter plan. Upgrade to the full plan when you're ready and apply it one tactic at a time.
Set it up in three commands
npx diffmode@latest logindiffmode skill installdiffmode run <product> --from-url https://your-product.comThat's the whole setup. The first run is your free audit — covered by the credit you get at signup.
Before your first run: you need a
free Diffmode account — the
browser login creates the token. Adding credits for the full plan is
browser-only: the CLI never takes
a card, it just points you to
diffmode.app/app/billing
when you're ready.
Two ways to drive it
Conversationally, with your agent
Install the skill and ask Claude Code, Codex, or Cursor in plain English — “run a growth plan for my-product.com.” Your agent handles sign-in, checks your credits, submits the run, watches it, and pulls back the results. This is the easy path, and the one most people use.
Raw CLI, for terminals and CI
Prefer the command line, or wiring it into a script? Run the commands
yourself. Pass
--json for
machine-readable output and lean on the exit codes — it's built to
run unattended in CI.
A free audit first — the full plan when you're ready
Every product starts with the free audit. Generic AI tools hand your agent the same playbook every competitor runs — Diffmode builds tactics you'd actually run, matched to your constraints. Get the full plan with credits once you've seen the depth.
Your free audit
- Landscape — your competitive field and where the gaps are.
- Advantages — the angles you can own that competitors can't.
- Buyers — who they are and what actually moves them.
- Blockers — the objections to clear before they buy.
- Growth plan — a starter set of tactics to begin with.
When you get the full plan
- 7–10 prioritized tactics, each traced to the growth mechanisms behind it.
- Day-by-day execution — what to do Monday morning, not a strategy deck.
- Pre-filled templates — ready to use, not fill-in-the-blank.
- Kill criteria — when to stop if a tactic isn't working.
- PDF export plus a stakeholder one-pager.
A full plan is dozens of files. The CLI pulls them on demand — a manifest first, then the file you ask for — so your agent reads what it needs instead of dumping the whole bundle into the conversation.
How this differs from the free plugin
Both run the Diffmode method. The free plugin shows you the method on a lighter database, on your own machine. The CLI runs the real thing.
| Dimension | Free plugin | Diffmode CLI |
|---|---|---|
| Where it runs | On your machine, inside Claude Code | Calls the hosted Diffmode API |
| Cost | Free and open-source | Paid credits — your first audit is free |
| Database | A lighter database, mined fresh each run | The full 576-mechanism database |
| How deep it goes | Stops at tactic ideas | Full plan — prioritized, day-by-day |
| Best for | Trying the method for free | Running the real Diffmode from your terminal or agent |
Want the free, local version first? See the free Claude Code plugin →
Start free, pay as you go
Your first audit is free — covered by the credit you get when you create an account. Top up with a credit pack — Starter, Builder, or Scale — whenever you're ready for the full plan. There are no subscriptions and no contracts — you only spend credits when you get the full plan.
Packs available: Starter, Builder, Scale. Billing is browser-only.
30-day money-back guarantee.
What you need to run it
Node 20 or newer
That's the only install requirement.
npx diffmode@latest
fetches the CLI for you — or
npm i -g diffmode
to install it globally.
A free Diffmode account
Sign in with the browser login. Your account comes with one credit — enough for your first free audit.
An AI agent (optional)
Claude Code, Codex, or Cursor lets you drive the CLI by chatting. Skip it and the raw CLI works on its own.
Credits for the full plan
The free audit is covered. The full plan costs credits — top up with a pack on the billing page.
Works with Claude Code, Codex & Cursor
diffmode skill install
sets up all three by default, so your agent knows how to sign in, check
credits, run the pipeline, and read the results. Want just one? Add a
--target flag.
diffmode skill install --target claudeFrequently asked questions
Is the Diffmode CLI free to start?
You can start free. Creating an account gives you one credit, which covers your first growth audit — the same free audit the web app includes. The CLI itself is free to install and run. You only spend credits when you get the full plan. Node 20 or newer is the only requirement.
How is this different from the free plugin?
The free plugin runs entirely on your own machine and builds a lighter database from public case studies each run, stopping at tactic ideas. The CLI calls the hosted Diffmode API — the same pipeline the web app runs — over the 576-mechanism database, and goes all the way to a prioritized, day-by-day plan. The plugin is free and open-source. The CLI is the paid product, with a free first audit.
Do I need Claude Code to use it?
No. The CLI works on its own from any terminal. Claude Code, Codex, and Cursor are optional — install the skill and your agent can drive the CLI for you, but you can run every command by hand if you'd rather.
What does a run cost, and how do credits work?
Runs are paid in credits. Your first audit is covered by the credit you get when you sign up. The full plan costs more credits, which you top up with a pack — Starter, Builder, or Scale — on the billing page. Billing is browser-only: the CLI never takes a card, it just points you to the billing page when you're low.
Where do I get a token?
The browser login handles it for you — npx diffmode@latest login opens a window and stores the token. If you're on a headless machine, or prefer to do it by hand, mint a personal access token at diffmode.app/app/tokens and pass it to diffmode login.
Can I script it or run it in CI?
Yes. The CLI is built for automation — pass --json for a stable, versioned response on stdout, with progress and errors on stderr. Exit codes are part of the contract, so a script or CI job can branch on sign-in, credits, conflicts, and rate limits without scraping text.
Which agents are supported?
Claude Code, Codex, and Cursor. diffmode skill install sets up all three by default. Add --target claude, --target codex, or --target cursor to install for just one.
Is it the same as the web app?
Yes — it runs the same hosted pipeline over the same 576-mechanism database, and bills from the same credits. The browser, the CLI, and your AI agent are just different front doors to one product. Start an audit in your terminal and the results show up in your account the same way.
Run your first audit from the terminal today.
Sign in, add the skill, and ask your agent for a growth plan. The first audit is free — credits only when you get the full plan.
npx diffmode@latest login