Buy Claude Pro & Max in Nepal — Claude Code, Cowork, Three Tiers | Cheapmandu
Upgrade to Claude Pro, Max 5x, or Max 20x. Full model lineup, Claude Code, Cowork, Skills, Claude Design, Claude Science. Pay with eSewa, Khalti, or Fonepay QR.
Most AI assistants optimise for a fast answer. Claude is built for work where accuracy and reasoning matter more than speed — long documents, real codebases, analysis you’ll act on.
The free tier is fine for trying it. If you’re using Claude for real work, the usage ceiling is what stops you, and that’s what Pro and Max remove.
Cheapmandu activates Pro, Max 5x, and Max 20x for Nepali users with local payment. No international card needed.
The Models
Anthropic ships roughly every four to eight weeks, so this section moves. Current as of early September 2026.
Claude Opus 5 (24 July) — the general-purpose flagship. Close to Fable’s capability at roughly half the price, and leading on coding and knowledge-work evaluations including Frontier-Bench and GDPval-AA. Anthropic describes it as its most aligned Opus model.
Claude Sonnet 5 (30 June) — the default on Free and Pro. Near-Opus quality, considerably faster, with strong gains over Sonnet 4.6 on multi-step reasoning, tool use, and coding. Most people work here and switch up when a task needs it.
Claude Haiku 4.5 — fastest and cheapest. Quick lookups, translations, high-volume light work.
The Mythos tier sits above Opus, and it moved on 1 September with Claude Fable 5.1 and Claude Mythos 5.1. Both are built for demanding reasoning, long-running agents, coding, multistep research, and document-heavy professional work.
The two share the same underlying model — Fable is the version carrying additional safeguards around biology, cybersecurity, and AI research. Fable is the one you can reach on a consumer plan: Max includes it up to 50% of your weekly limits, while Pro accesses it through usage credits. That’s the clearest single reason to choose Max over Pro.
Because of those safeguards, a small share of Fable queries get answered by Opus 5 instead. Anthropic tuned this conservatively and says it triggers in under 5% of sessions.
One useful behaviour new in 5.1: when Fable is working through a long agentic task, it now writes short progress updates between tool calls rather than going silent. On a task that runs for several minutes, you can see what it’s doing instead of watching a spinner.
Claude Mythos Preview isn’t publicly available — it’s with a small number of organisations under Project Glasswing, which uses it to scan critical software for security vulnerabilities.
If you’re researching Fable: Anthropic suspended Fable and Mythos on 12 June to comply with US Department of Commerce export controls, and restored both on 1 July once those lifted.
What Opus 5 Changed
The effort ladder. Opus 5’s main control is how much computing effort it spends, across five settings — low, medium, high, xhigh, max. Lower effort keeps much of the performance while using far fewer tokens.
On a subscription this is the single most useful thing to understand, because effort burns your usage allowance. Running routine work at low effort leaves headroom for the problems that actually need it. It’s the main lever you have for stretching a Pro plan.
Fast mode — roughly 2.5x the default speed.
Mid-task model switching — change model partway through a task. Start on Haiku, escalate when it gets hard.
Better self-correction — markedly stronger at checking its own work and iterating until it succeeds. Fewer rounds of “no, try again.”
It’s also Anthropic’s most capable generally available model for scientific research, and notably stronger on biology than its predecessor.
Recent Features
Cowork now runs everywhere, and keeps working when you close the laptop
This is the biggest change since the summer. Cowork started as a desktop feature; as of September it runs on web and mobile as well, with sessions, files, and state synced to your Claude account across devices.
The part that matters: remote execution is now the default. The agent loop and any code it runs happen inside isolated cloud sandboxes on Anthropic’s infrastructure, with no outside network access unless it’s explicitly allowlisted and file access limited to folders you authorise. Two practical consequences:
- Work continues after you shut your laptop. Start a long task, close the lid, check the result on your phone.
- Scheduled tasks run server-side with no device online at all.
For anyone on an unreliable connection or working across a laptop and a phone, that’s a meaningful shift — a dropped connection no longer kills a running job.
Chat and Cowork now also share a single home for Projects and Artifacts, so work started in one surface is available in the other.
Skills
Skills are reusable instruction packages Claude loads when a task calls for them. Anthropic ships pre-built ones for Excel, PowerPoint, Word, and PDF work, and they’re composable — Claude coordinates several on its own without being told which to use.
They run across the Claude app, Claude Code, and the API, so a skill you rely on behaves the same wherever you’re working. If you produce the same kind of document repeatedly — reports in a house format, decks to a template — this is the feature that makes it repeatable.
Memory works everywhere, and you control it
Anthropic merged the memory systems behind chat and Cowork in late August. Claude carries what it learned in one place into the other, so you stop rebriefing an agent on things you already told it in conversation.
The more useful half: you can read, edit, and delete what Claude has retained, topic by topic. If you’ve avoided persistent memory on privacy grounds, that’s the change worth knowing about.
Claude Science
A research workbench rather than a chat feature, launched 30 June. It ships with 60+ curated skills and connectors preconfigured for genomics, single-cell analysis, proteomics, structural biology, and cheminformatics, produces auditable artifacts, and scales from a laptop up to HPC clusters while keeping sensitive datasets on local infrastructure. Pro and above, on macOS and Linux.
Before you buy for research work: Anthropic has expanded its support for scientists with 10,000 free and discounted team seats and broader AI for Science credits for researchers on high-impact projects. If you’re at a Nepali university or research institute, check whether you qualify before paying anyone — including us.
Connectors
Claude lists over 950 MCP servers in its connector directory. The July MCP spec update brought a stateless core, stronger OAuth and OIDC authorisation, embedded UI in connectors, and private network tunnels, which mostly matters because connectors became more reliable and easier for services to ship.
Practically: Claude reaches into Drive, Slack, GitHub, Zoom and hundreds of other tools without copy-paste. Salesforce shipped a dedicated plugin in late August with 37 prebuilt sales skills, if that’s your world.
Claude Code
The terminal and IDE agent, updated almost weekly. Recent additions include a built-in Concise output style that leads with results and skips narration, a SendFeedback tool that drafts a report when something goes wrong in a session, cost optimisation for API spend, much larger inline output limits before results get written to a file, a default-model setting, GitLab merge request support, and screen reader support in VS Code.
Pro and above; not on free.
Everything else
- Claude Design — describe a landing page, deck, or UI and get a first visual you refine in conversation.
- Claude for Excel, PowerPoint, Word, and Outlook — Claude inside Office rather than beside it.
- Claude in Chrome — a browsing agent that reads and acts on pages.
- Claude Tag — tag @Claude in Slack to hand off a task.
- Claude Academy — free learning materials on working with AI, much of it deliberately model-agnostic rather than Claude-specific.
- Unlimited projects, Research, Artifacts, file creation.
Two things worth knowing
How the limits actually work. There are two: a rolling five-hour window and a seven-day weekly cap that resets on a fixed day. Anthropic permanently doubled the five-hour limits on 6 May and removed peak-hour throttling for Pro and Max. It also runs temporary boosts from time to time — those expire, so don’t plan around a number you saw in the app last month.
Outputs are watermarked. Anthropic signed the EU Code of Practice on Transparency of AI-Generated Content in July and watermarks Claude’s text output globally. Images and video Claude produces through the code execution tool now also carry C2PA Content Credentials. None of it identifies you, your organisation, or your conversations — it marks the content as machine-generated. Worth knowing if you publish work made with Claude.
Three Tiers
| Tier | Usage | Official | Best For |
|---|---|---|---|
| Pro | Baseline (~5x free) | $20/mo | Individuals, students, writers, most developers |
| Max 5x | 5x Pro | $100/mo | Daily heavy use, professional developers |
| Max 20x | 20x Pro | $200/mo | All-day agentic coding, full-time AI workflows |
All three unlock the same features and models. The difference is capacity — plus Fable inclusion on Max.
Anthropic doesn’t publish exact message counts, because consumption depends on context length, model choice, attachments, effort setting, and tool use. Community testing puts Pro at roughly 10–45 prompts per five-hour window, Max 5x at 50–225, and Max 20x at 200–900. Treat those as rough — long documents and agentic coding burn far more than chat.
Who Actually Needs Max
Most people don’t. Max earns its price when:
- You run Claude Code for hours a day
- You consistently hit Pro’s weekly cap
- You want Fable included rather than metered
- Claude is central to your full-time work
If you use Claude for occasional writing, research, or study, Pro is the right answer and Max is money wasted.
Start with Pro. Track a month. Upgrade only if you actually hit the ceiling.
Price in Nepal
| Tier | Duration | Price (NPR) | Password? |
|---|---|---|---|
| Claude Pro | 1 month | Rs. 4,500 | No — email only |
| Claude Max 5x | 1 month | Rs. 17,000 | No — email only |
| Claude Max 20x | 1 month | Rs. 25,000 | No — email only |
Being straight about the value
Pro at Rs. 4,500. Official is $20, around Rs. 2,760. Even allowing for card fees and conversion, this is a premium, not a discount. What you’re paying for is the ability to pay at all — eSewa or Khalti instead of a dollar card — plus local support. If you have a working international card, subscribing directly is cheaper.
Max 5x at Rs. 17,000. Official $100 is roughly Rs. 13,800. Same situation: convenience, not savings.
Max 20x at Rs. 25,000. Official $200 is roughly Rs. 27,600, so this is modestly cheaper than direct — around Rs. 2,600 a month.
We’d rather say this plainly than dress it up. If you can pay Anthropic directly, do. This service exists for people who can’t.
How Activation Works
- Choose your tier.
- Give us the email on your Claude account. That’s all we need — we never ask for your password.
- Pay via eSewa, Khalti, or Fonepay QR.
- Your access is set up, usually within 30 minutes, occasionally up to 24 hours.
Please read before buying: access is provisioned through an organisation account, not added to your existing personal subscription. Your work lives in the workspace we set you up in, and your existing personal chats and projects are not automatically part of it. Ask us before you buy if you have existing work you need to keep using — we’ll tell you exactly what carries over.
Access is monthly and prepaid. When it ends, your account returns to free tier limits. Your data stays where it is.
Build Around Claude
GitHub Copilot in Nepal — Copilot for in-editor completions, Claude Code for whole tasks. Different jobs; most developers run both.
Microsoft 365 Personal in Nepal — Claude for Excel and Word, and the Office Skills, need Office underneath them. Rs. 2,999/year.
CapCut Pro in Nepal — Claude writes and structures the script; CapCut does the edit.
Paperpal AI Pro in Nepal — journal formatting and citation compliance after Claude drafts the paper.
Higgsfield AI in Nepal — Claude doesn’t generate video or images. This does.
FAQ
Do you need my password? No. Only the email on your Claude account.
What’s the difference between the tiers? Same models, same features, different capacity — roughly 5x free on Pro, 5x Pro on Max 5x, 20x Pro on Max 20x. Max also includes Fable within your weekly limits rather than metering it.
Which should I start with? Pro. Most people, including daily users, never exceed it. Move up only after you’ve hit the ceiling repeatedly.
What models do I get? The current lineup — Opus 5, Sonnet 5, Haiku 4.5 — on all tiers, plus new models as Anthropic releases them. Fable 5.1 is included on Max; on Pro it runs through usage credits.
What’s the effort setting for? It controls how hard Opus 5 works on a task, from low through max. Lower settings use fewer tokens, so on a subscription they stretch your allowance further. Save the higher settings for problems that need them.
Can Cowork keep running if I close my laptop? Yes. Cowork sessions run remotely by default in isolated cloud sandboxes, so a long task continues after you shut the lid and scheduled tasks run with no device online. It works on web and mobile as well as desktop.
What are Skills? Reusable instruction packages Claude loads when a task needs them, with pre-built ones for Excel, PowerPoint, Word, and PDF. They work the same across the Claude app, Claude Code, and the API.
How do the usage limits work? Two limits run together: a rolling five-hour window and a weekly cap that resets on a fixed day. Anthropic doubled the five-hour limits in May.
Does Pro include Claude Code? Yes. Claude Code needs at least Pro — it isn’t on the free tier. Max gives you a lot more of it.
Is Claude Science included? Yes, on Pro and above, on macOS and Linux. If you’re a researcher, check Anthropic’s AI for Science programme first — there are free and discounted seats for qualifying projects.
Can I control what Claude remembers? Yes. Memory spans chat and Cowork, and you can read, edit, or delete what it has retained on any topic.
Will my existing chats and projects carry over? Ask us first. Access is provisioned through an organisation workspace, so what transfers depends on your setup. We’ll tell you before you pay.
What happens when the month ends? Your account returns to free tier limits. Your data isn’t deleted. Renew when you want.
Can I upgrade or extend mid-cycle? Message WhatsApp with your email and current tier. Usually possible, prorated.
How fast is activation? Usually 30 minutes, occasionally up to 24 hours.
Payment methods? eSewa, Khalti, Fonepay QR.
Why Cheapmandu
- No password, ever — email only
- Local payment: eSewa, Khalti, Fonepay QR
- All three tiers, so you can right-size instead of overbuying
- WhatsApp support in Nepali
- Honest about pricing: Pro and Max 5x are convenience, not discount
- Operating since 2024
Pro Rs. 4,500 · Max 5x Rs. 17,000 · Max 20x Rs. 25,000 — per month
Reviews
There are no reviews yet.