Human First. AI Assisted.

Set up Claude the right way.

ClevSkill is the managed safety layer for Claude Code: one installer provisions your desktop with a smart safety guard , an encrypted key vault, 86 curated skills, an 8-area weekly health scan, and token cost controls. Non-engineers get Claude that works and can't delete their files. Teams and MSPs get standardized governance across every seat. From $7 a month; consultants charge hundreds for a one-time setup that goes stale.

Windows 10 and 11No terminal, everYou approve every changeKeys never leave your machine
Who it's for

Pick your path

minutesto a fully governed setup
0files lost to guard-blocked actions
20active security skills included
8 areasDoctor scans weekly, per desktop
12,000+US MSPs who need this today
The problems

Sound familiar?

Real things real Claude beginners say. Each one is something Clevskill was built to end.

Everyone says use Claude. Nobody says how to set it up.

Set up the right way, guided

A friendly installer asks a few questions and configures everything the way an expert would: the right skills, memory files, and settings. No terminal, no config files, no guesswork.
I'm scared it will break something or delete my files.

Your files are backed up before anything runs

A guard checks every command the AI wants to run. Risky operations are blocked; ordinary deletes are backed up first with one-click restore. Everything is logged so you can see exactly what happened.
I installed skills from the internet and half don't work.

Curated skills that actually run

Most marketplace skills fail quality checks. Yours come pre-tested and health-checked, with a doctor that catches breakage before you hit it.
My Claude plan keeps running out mid-week.

It runs out because it wasn't set up right

A bloated setup silently re-reads the same files every session and burns your plan. Clevskill tunes your memory files and context for token efficiency, shows you live what each project costs, and warns you at 50% and 80% of your budget.
I only get value when I'm sitting there typing at it.

Put Claude to work on a schedule

Morning inbox triage, weekly reports, content pipelines: Clevskill sets up scheduled automations (cron) and Cowork assistants that run approved recurring jobs, so the work happens whether you're at the desk or not.
It worked last week. An update broke it and I can't debug this.

Maintained for you, automatically

A weekly health scan catches drift, one-click repair fixes it, and nothing changes without your approval. Your setup stays working without you babysitting it.
I got a new machine. Now I have to rebuild everything from scratch.

New machine ready in minutes

Run one command on the old machine to export an encrypted profile: all your skills, settings, MCP configs, and vault key references. Import it on the new machine. Done. Multiple desktops stay in sync from one profile update.
My API keys are in .env files and old chat logs. I know that's a problem.

Every key encrypted, vault-locked

Clevskill stores every API key in an AES-256-GCM vault, with the master key tied to your OS keychain. No plaintext .env files, no keys accidentally pasted into chat history, nothing to leak. One exposed .env file can turn into a five-figure surprise bill.
I have no idea what the AI is accessing or sending. That worries me.

Complete local audit trail, nothing hidden

Every command the AI runs, every file it touches, every key it uses: logged locally with a timestamp and decision reason. Nothing leaves your machine without you seeing it first. The audit log is yours, not ours.
The skills

Seven arsenals, one install

86 pre-tested skills land on every desktop, grouped into what they do for you.

The Doctor

Your Claude setup has a doctor now

The Doctor examines eight areas of your installation: skills, configs, MCP connections, memory bloat, secrets posture, usage costs, version drift, and safety hooks. It shows you what it found in plain language, prescribes the fix, and applies it only with your approval, after taking a backup. It runs weekly on its own, or whenever you ask.

This is what a consultant would bill hours for, built in and always on call.

you: my setup feels broken since the update
doctor: scanned 8 areas. 2 must-fix:
  1. mcp server "playwright" not responding
  2. CLAUDE.md grew 412 lines (token waste)
fix both? [backup first, reversible]  y/n
Security at a glance

Locked down by default

Blockedcatastrophic commands never run
Backed upbefore any delete; one-command restore
Scannedemail links, attachments + agent browsing
Encryptedkeys in a vault, never plain files
Loggedcomplete local audit trail
Collision-checkedsame-name keys across projects flagged

Full mechanics, monitored metrics, and how the Vault works with your tools: the security deep-dive.

What's included

Everything you get

Not a marketplace, a managed setup. Every piece below is included, configured, and kept healthy for you. Want the exact mechanics? See guardrails, monitoring metrics, and the Vault, or the step-by-step training guides.

86 curated skills, pre-tested

A security suite (email, link, and attachment scanning, prompt-injection defense, secret-leak prevention), setup-and-maintenance skills, document and spreadsheet tools, business coaching and advisors, and creator tools for photos, video, music, and life. Every one health-checked before it ships.

Everything works on day one; nothing from the 73%-fail marketplace.

Guard + backup + complete audit log

Every command is checked before it runs. Catastrophic operations are blocked outright. Ordinary deletes are backed up first with one-command restore. Every decision is logged with a timestamp. Notably: Claude Cowork activity is not in Anthropic's enterprise audit logs. Ours is.

Let the AI work without watching it, and prove what it did.

Token optimization built in

Trimmed memory files, lean CLAUDE.md, no context bloat re-reading the same files every session. When 'my plan keeps running out' is really 'my setup is wasteful', this is the fix.

The same Claude plan lasts noticeably longer.

Stay inside your plan

A live cost dashboard prices your real usage by project and model, with alerts at 50% and 80% of your weekly budget and a hard look at what's eating it.

No more mid-week surprises or mystery burn.

Claude works while you don't

Scheduled (cron) automations and Cowork assistants for the recurring work: morning email triage, weekly status reports, content pipelines, monitoring sweeps. Always with your approval rules attached.

Hours back every week, on a schedule, not just when you're typing.

Encrypted vault + clevskill-run launcher

Every API key lives in an AES-256-GCM vault scoped to four levels: personal, project, client, and company. The master key stays in your OS keychain. The clevskill-run launcher injects vault keys into any process at startup (VS Code, GitHub CLI, n8n, anything), so nothing sits in a plaintext file and nothing leaves your machine.

A leaked laptop file can't become a breached account.

Machine migration + employee handoff

Three modes: clone-exact copies your full setup to a new machine in minutes; employee-handoff scrubs personal data before handing a machine to a new hire; team-baseline rolls one approved standard to every managed seat from the portal. Export is encrypted, passphrase-protected, and never bundles plaintext secrets.

New machine in minutes. Offboarding without data residue. Every seat identical.

Email links + attachments, scanned

If Claude works your Outlook or Gmail, every inbound link and attachment is scanned BEFORE anything acts on it: local phishing heuristics plus VirusTotal threat intelligence, failing closed when in doubt. Verdicts cache for 24 hours, so repeat checks are instant and don't burn your scan quota.

The AI never clicks what you wouldn't.

Safe agent browsing

When the AI browses with Playwright or researches with Firecrawl, every URL is vetted before it's visited. Malicious destinations are blocked outright; suspicious ones get flagged into the conversation so you decide.

Web automation without wandering into traps.

The Doctor: weekly scan + repair

Eight-area health scan every week (skills, configs, MCP servers, costs, secrets posture, version drift), one-click repair, and self-healing that always asks before changing anything.

It keeps working after updates, without you babysitting it.

Two-track deployment: Claude Code and Claude Team app

ClevSkill handles both Claude surfaces differently. For Claude Code users, safety rules are enforced by hard hooks that run before every command. For Claude Team app (Cowork) users who have no hook system, ClevSkill generates behavioral guardrails via Project instructions and a mirrored audit trail. No competitor addresses this split. Both tracks get the vault, the Doctor scan, and fleet visibility.

One product governs both Claude surfaces; no seat is left unmanaged.

Cross-project secret collision detection

If the same key name exists in more than one project with different values, or if the same secret is duplicated across projects where it could be consolidated, the Doctor flags it. Eliminates the 'wrong key in the wrong project' failure mode that causes silent, maddening bugs in multi-project setups.

Catch key conflicts before they cause production failures.
How it works

Working in three steps

Sign up (2 minutes)

Tell us what you're setting up. Pick a plan sized to your desktops; your license issues the moment checkout completes.

Run the installer (minutes)

Download ClevskillSetup, run it once per desktop, paste your enrollment code. The wizard does the rest and verifies everything works.

Relax (automatic)

Weekly health scans roll up to your dashboard. Problems get flagged and repaired, with your approval, before they bite.

Pricing

Simple per-desktop pricing

A seat is one installed desktop. Launch offer: all plans 50% off, locked in for as long as you stay subscribed. Nonprofit? Email us for 40% off. MSP / partner? Wholesale pricing is here. 250+ seats? Talk to us about Enterprise.

$150One-time onboarding, per desktop. Guided setup of each machine, safety rails wired, everything verified working before you pay a second month. Drops to $75 with annual prepay.
$0See what Claude actually costs you, and cap it. No card required.Cost dashboard on your machine + weekly-cap alerts · Free baseline skill pack via the installer · Community support · Upgrade anytime, keep everything
Start free, no card

Solo

$14$7/mo
50% launch offer
One desktop. A third of your Claude subscription to keep all of it working safely.
  • Guided setup included
  • 86 curated skills (incl. the 20-skill security suite)
  • Safe-ops guard + backup + one-click restore
  • Encrypted key vault (4 scopes, OS keychain)
  • Token optimizer + plan-cap alerts
  • The Doctor: 8-area weekly scan + repair
  • clevskill-run: vault keys into any tool
  • Fixes when Claude updates break things
  • Email support

or create your account first

Personal Sync

$24$12/mo
50% launch offer
Up to 3 of your own machines: copy your full setup between desk, laptop, and home.
  • Everything in Solo, on 3 machines
  • Encrypted secrets bridge: keys travel safely, never plaintext
  • Same skills, memory, projects on each machine
  • Move to a new computer in minutes

or create your account first

Business

$28$14/seat/mo
50% launch offer
51-250 seats. Audit logs, SSO prep, custom skill approval. ACH/invoice billing.
  • Everything in Team
  • clevskill-mcp health + self-heal
  • SSO prep
  • Custom skill approval workflow
  • Volume pricing at 100+ seats
  • Dedicated support

or create your account first

What will it cost me?

Teamplan
$10per seat/mo
$50per month
$500per year (2 months free)

A seat is one installed machine. Compare: your Claude subscription alone is $20+ per person per month; clevskill keeps all of it safe for a fraction of that.

Setup only vs. managed

Setup is included free with every subscription. The one-time option exists for people who want to maintain their own setup.

Setup only ($150 once)Managed (from $5/mo with launch offer)
Guided install + verification✓ included free
86-skill curated pack + safety guard + vault✓ (as of install day)
Fixes + updates when Claude changes break things30 days✓ continuous
The Doctor: weekly scan + one-click repair
Monitoring dashboard + alerts
Copy your setup between computers
Seat management + support30 days

Either way, what's installed is yours: cancelling never removes skills or your files. Maintenance, updates, and monitoring are what the subscription buys.

If you cancel, in plain terms

You keep, forever

  • Every installed skill, as it is that day
  • All your files, projects, and history
  • Your keys (the vault stays readable + exportable)
  • Your Claude subscription, untouched

What stops

  • Fixes + updates when Claude changes break things
  • The Doctor: weekly scans + one-click repair
  • Monitoring dashboard + alerts
  • Copying your setup to new computers + seat management
  • Support

Setup only (no subscription)

$150/desktop, one-time
Today's pack installed, verified, and yours to keep, with 30 days of support. No ongoing updates, scans, monitoring, or sync; you maintain it yourself.

White-glove setup

$497/one-time
We install and configure remotely, up to 5 machines, and hand you a verified working setup.
FAQ

Questions

I'm brand new to Claude. Is this for me?

Yes, that's exactly who we built it for. You bring a Claude subscription from Anthropic; Clevskill sets up your desktop the way an expert would and keeps it that way. You never touch a terminal.

I already have my own skills and customizations. What happens to them?

They stay, and they get covered. Setup starts with Detect, which inventories what's already on the machine. Your own skills get the same weekly health checks ours do; your configs, memory files, and key storage get scanned for problems; and nothing changes without a backup snapshot first and your approval. If something unusual in your setup is intentional, mark it as an exclusion and it stops being flagged, visibly, never silently.

I already have Claude set up and it's a mess. Can you fix it?

Yes. The installer has a Repair mode: a read-only diagnostic first scans eight areas (skills, configs, MCP servers, memory bloat, secrets posture, costs, version drift), shows you exactly what's wrong, and fixes it with your approval. Your files, projects, and history stay untouched; a backup is taken before any change.

What exactly does Clevskill manage?

Your Claude Code and Claude Cowork desktop installations: setup, curated skills, safety guards, encrypted keys, usage-cost hygiene, and ongoing health monitoring. We never proxy or resell tokens.

How does ClevSkill protect my files?

Every command the AI runs passes through the guard: catastrophic operations are blocked outright, ordinary deletes are snapshotted to a backup folder first with one-command restore, and every decision is logged with a timestamp. The audit log shows exactly what the AI did and didn't do. The guard significantly reduces the risk of accidental file loss, though no software can guarantee zero risk from all possible actions.

Doesn't Claude already protect my files?

Partially. Claude Cowork asks before permanently deleting files, which is good. ClevSkill goes further: we back up BEFORE any destructive operation runs (so mistakes are reversible with one command, not just refusable), and we keep a complete local audit log of everything the AI did. Notably, Cowork activity isn't included in Anthropic's enterprise audit logs; ours is.

How fast is setup, really?

Minutes per desktop. The installer detects your current setup, installs your pack, wires the safety hooks, and verifies the result before it says done.

What counts as a desktop (seat)?

One installed machine. Your license encodes the seat count; add or remove desktops any time from your dashboard.

Why pay monthly if the skills are installed once?

The subscription isn't access to skills, it's keeping them working. Claude updates regularly and setups break; subscribers get continuous skill updates, the weekly health scan, one-click repair, the monitoring dashboard, sync between computers, and support. A setup-only purchase is yours forever, but it's frozen on install day and you maintain it yourself.

What happens if I cancel?

Your license simply expires at the end of its period. Your installed skills, files, and keys remain on your machine. What stops is what the subscription provides: updates, scans, repair, monitoring, and sync.

Where do my secrets and license live?

In an encrypted vault on your own machine, with the master key in your OS keychain. A copy of your license also lives on your account page. Neither is ever sent anywhere else.

What does monitoring send you?

Only the scoreboard: finding counts and titles from the weekly health scan, engine version, and a timestamp. Never file contents, configs, keys, or chat history. It is opt-in at enrollment.

Stop fighting your setup. Start using it.

Get set up in minutes