How to install the Claude desktop app and choose the right plan — whether you're setting up for yourself, a finance team, or your whole organisation. Plain English, no IT project required.
The fastest way to get going. Install the desktop app, sign in, and start on the Free plan — upgrade to Pro when you want Claude Code and Cowork.
Best for: trying Claude out, a single Finance Director, or a personal proof of concept before rolling it out wider.
Go to claude.ai/download and download the app for your computer (Windows or macOS). You can also use Claude in any browser at claude.ai, but the desktop app gives you the smoothest experience and access to Cowork.
Run the installer you've just downloaded and open the Claude app. On Windows it installs in a couple of clicks; on a Mac, drag Claude into your Applications folder.
Sign in with your email, Google account, or Apple ID. If it's your first time, you'll create an account in under a minute and land on the Free plan.
Free is enough to explore. To unlock Claude Code and Cowork — the tools we use to build finance automations — upgrade to Pro. Max adds substantially more usage for heavy users. Compare everything on the plans & pricing page.
There's no nonprofit discount on the Pro or Max plans — the charity rates only apply to Team and Enterprise. So if your goal is to get Cowork and Claude Code at the lowest cost, two Team seats often work out cheaper than one Pro licence, and you also get a Data Processing Agreement. See the Team option below.
One shared workspace with central billing, admin controls, and the option of single sign-on. Everyone gets Claude Code and Cowork, and your data isn't used to train models by default.
Best for: a finance team or a small-to-medium charity rolling Claude out to several people.
From claude.com/pricing, choose Team and set up billing. One person becomes the workspace owner/admin — usually the person managing the rollout.
Name your organisation's workspace and set the basic preferences — including the data setting that keeps your content out of model training (on by default for Team).
Add colleagues by email from the admin console, or connect single sign-on (SSO) if your organisation uses Microsoft or Google identity. Each invitee gets a join link.
Each team member downloads the app from claude.ai/download and signs in with their work account — not a personal one — so they join the shared workspace.
As admin, you manage seats, billing, and usage centrally, and can adjust data and sharing settings for the whole workspace from one place.
Registered charities qualify for discounted Team seats with a two-seat minimum, eligibility verified through Goodstack. Team includes Cowork and Claude Code, keeps your data out of model training by default, and a Data Processing Agreement is available (it isn't on Pro or Max).
Pricing changes from time to time and is set in US dollars, so confirm the current charity rate on the pricing page before you commit. Happy to help you scope the right number of seats — just book a call.
For larger organisations that need automated user provisioning, single sign-on, audit logging, role-based access, and stronger compliance — arranged with Anthropic's sales team rather than self-serve.
Best for: larger charities and organisations with IT governance, security, or regulatory requirements (including a HIPAA-ready option).
Enterprise is arranged directly with Anthropic. Start from the pricing page and choose Contact sales on the Enterprise plan to discuss seats, pricing, and your requirements.
Your IT team sets up SSO and SCIM, so staff sign in with their existing work credentials and accounts are created and removed automatically as people join or leave.
Set up role-based access, audit logs, data-retention rules, and spend controls centrally. The compliance API and audit logging give your board and auditors the assurance they need.
Staff install the app from claude.ai/download (or your IT team deploys it centrally) and sign in via SSO to join the managed organisation.
Enterprise customers can layer on additional capabilities such as Claude Security. Your Anthropic contact will help you scope what's relevant to your risk profile.
Before anyone sets anything up, these two come up. Both are answered properly in under five minutes, by Anthropic’s education team rather than its marketing — and neither video is selling anything.
The video splits it into four separate uses. That turns “is this safe?” — unanswerable, and therefore a stall — into four questions a finance team can actually answer.
In charity finance, almost none of the data is ours. It belongs to donors, beneficiaries and staff, which changes who the question is really being asked on behalf of.
Watch “What does AI know about me?” →Its central idea is that trust should be a dial rather than an on/off switch. How hard you check depends on what the answer is going to be used for.
Finance invented that dial a long time ago and already has a word for it. It is materiality, and it is a judgement every qualified accountant has been making for years.
Watch “Can you trust what AI tells you?” →Both videos are Anthropic’s own and open on YouTube. We link them because they answer the questions well — neither one mentions a plan or a product.
Most of the difference between an answer you can use and an answer that reads like a textbook comes from what Claude knows about you before you ask it anything. It has read a great deal about charity finance in general. It knows nothing about your funds, your cost centres, your year end, or the fact that your Finance Committee only wants three numbers on the first page. Four things are worth getting right at the start.
There is a settings area for instructions that apply to every conversation, and it is worth a look. We lead with files for a practical reason rather than a technical one.
A file can be shared with a colleague, reviewed by your manager, kept under version control, and inherited by whoever does the job after you. A personal setting can do none of those, and finance work is rarely a one-person concern.
Standing context is read every time you work, so treat it as something that persists rather than a conversation you can delete.
No donor, beneficiary, patient or employee names. No individual records of any kind, no bank details, no credentials. Structure, conventions, definitions and preferences are what belong there — and keeping personal data out is far easier than taking it out later.
A skill is one task you have taught it: the monthly pack, the bank reconciliation, the quarterly VAT return. It carries the steps, the file names and the checks for that piece of work.
The standing context sits underneath every skill you will ever save. It is the reason the second one takes less time to set up than the first.
This is the part that decides whether any of it survives, and it is the part almost everyone drops.
Context goes stale quietly. A fund closes, a system changes, your reporting moves on — and because the output still looks plausible, nothing announces that the rules no longer describe how you work. It drifts back towards generic, and it is very hard to say when that started.
Claude offers more than one place for standing context to live. It can sit alongside your data in a folder you point Claude at for a particular piece of work, or it can be held in the workspace so that it is there permanently.
Those are different data-processing arrangements, and which one is appropriate depends on your own data protection impact assessment and what your DPO has agreed. It is a decision worth settling before you set anything up, rather than discovering it afterwards.
A quick comparison. Figures are indicative and in US dollars — always confirm the current rate on the official pricing page.
Indicative pricing as at June 2026. Anthropic updates its plans and prices regularly — check the live figures at claude.com/pricing. Charity (nonprofit) rates apply to Team and Enterprise only, with a two-seat minimum, and eligibility is verified through Goodstack.
If you'd like a hand choosing the right plan, getting set up, or building your first finance automations, we're happy to help.