Your users say what they want.
Your product does it.
Wizardful is an AI copilot you drop into your SaaS with one script tag. It answers questions from your docs, walks users through a task with a narrated ghost cursor, or just completes the task itself — in your real UI. No action wiring. No SDK work. Your existing interface is the API.
Free forever tier · 50 runs a month · no credit card
New contact
Type it. Watch it happen.
This is a real CRM with the real widget in it — the same one script tag you'd paste. Nothing here is a video.
New contact
That took ~13 seconds and one agent run. Ask again with a different name — it finishes in 0.7 seconds, with zero AI calls, and it costs us nothing. That gap is the whole product.
Now imagine this in your product. It's one line of HTML.
If it renders in a browser, it works
Ask. Show. Do.
The same assistant, at three levels of trust. You decide how far it goes — per task.
It answers, with receipts.
Wizardful crawls your docs and help centre and answers grounded in them, with a citation back to the page it came from. Then it does the thing every help widget stops short of: it offers to do it for you.
It moves the cursor. The user learns.
A labelled ghost cursor moves across your real UI and narrates each step, without touching anything. Onboarding, training and accessibility in one toggle. It is magical, and it cannot break anything.
It finishes the job.
The agent fills the form, flips the setting, creates the report — in the live UI, with an edge-glow so the user can see it working. You turn it on per task, once you have watched the skill do it right.
Show-me is on from day one. Do-it is earned, task by task, by you.
Live in five minutes.
Four steps, none of which involve your engineers.
- 1
Paste one tag.
That is the integration. There is no step where you register actions, wire callbacks, or expose an API.
<script src="https://cdn.wizardful.ai/w.js" data-key="wf_pk_…"></script> - 2
Point it at your docs.
Give it your help centre URL. It crawls, embeds, and grounds every answer in your own words.
- 3
Teach it the tasks that matter.
We read your docs and propose the top tasks your users will ask for. You watch it do each one, once. Each one becomes a skill.
- 4
Go live warm.
Your users arrive to a copilot that already knows your product — not one learning it on their time.
It gets faster and cheaper the more it's used.
Every other AI widget charges you the same for the ten-thousandth answer as the first. Ours stops charging.
The first time a user asks for something new, the agent works it out live by reading your UI. That takes about 13 seconds. Then Wizardful distils that run into a deterministic skill — a replayable recipe for that task in your app.
Every user after the first gets that skill instead of the agent. It re-checks its own work as it goes, and if your UI has changed under it, it falls back to the agent, solves it again, and recompiles itself.
And when your UI changes: the skill's own checks stop passing, so it falls back to the agent, solves the task against the new UI, and recompiles itself. The thing that silently kills a hard-coded product tour is, here, just a recovery path.
The agent reads your live UI and works the task out from scratch.
The distilled skill replays deterministically. No agent involved.
Zero AI calls. The expensive part happened once, ever.
Skills are unlimited on every plan, including free. They make the product cheaper to run — we're not going to charge you for them.
The question is "what if it deletes something?" Here’s the answer.
The #1 objection is the #1 feature. Nothing here is a roadmap item.
Show-me is the default.
A brand-new install cannot change anything in your product. Autonomy is something you switch on, per task, after you have seen the skill work.
Destructive actions are denied by default.
Delete, pay, send: blocked unless you allow them, and confirmed with the user even then.
You draw the boundaries.
Choose the pages and elements the agent may touch. Everything else is off-limits.
Every session is on the record.
Full step-by-step replay of what the agent saw, decided, and did. Audit it, or hand it to a customer who asks.
A link can never act silently.
Skill links your support team sends default to guided mode, and do-mode from a link always asks the user first.
You are the vendor here.
Browser agents act on the user’s side, uninvited. Wizardful is the layer you install, brand, constrain and audit.
Every question is a roadmap.
Wizardful sees what users try to do and where they get stuck — in their own words, at the moment they're stuck.
"142 people asked how to export to CSV this month." That's not a support statistic, it's a product decision.
You get the questions, the tasks users asked for, and the runs that failed — which is the one report no analytics tool can give you, because analytics only sees what users managed to do.
- how do I export to CSV?142
- change my billing email98
- bulk-assign owners to dealsruns failed74
- connect our Slack workspace51
- delete a workspaceruns failed33
Why not the thing you already looked at?
vs. product tours
Appcues, Userflow, Pendo, ChameleonTours are hard-coded to CSS selectors. One redesign and your onboarding silently dies — and you are paying $249/mo (Appcues) to $47k/yr (Pendo) for the privilege of rebuilding it. Wizardful works out the steps at runtime against whatever your UI looks like today. Tours that build themselves and never break.
vs. AI help chat
Intercom Fin, ZendeskThey answer. They cannot act. Fin charges $0.99 every time it resolves something — a meter that runs faster the better your product does. Wizardful resolves by doing, on a flat plan you can predict.
vs. copilot frameworks
CopilotKit, OpenCopilotPowerful, and they need an engineer to hand-wire every single action the copilot is allowed to take. That is the integration project you were trying to avoid. Wizardful needs zero wiring — it reads the UI you already shipped.
vs. building it yourself
page-agent is MIT, after allYou can — and our free tier is the DIY path. You would be signing up to run model keys, docs ingestion, guardrails, skill distillation and analytics forever, in exchange for the feature you were not planning to build this quarter.
Side by side
| Feature | Wizardful | Product tours | AI help chat | Copilot frameworks |
|---|---|---|---|---|
| Acts in your real UI | ||||
| Survives a redesign | ||||
| Answers from your docs | ||||
| Engineering to add an action | None | Author each tour | n/a | Wire every action |
| Setup | One script tag | Days | Hours | Weeks |
| Gets cheaper with use |
Priced like a tool, not a headcount.
A run is one completed interaction — an answer, a walkthrough, an action, a replay. Runs are pooled across every project in your org. Skills are unlimited on every plan.
Free
Enough to prove it works in your product.
Start free- Runs / month50
- Projects1
- SkillsUnlimited
- Ask, Show me, Do it
- Guardrails & session replay
- Remove Wizardful badge
- SupportCommunity
Starter
For a product finding its first thousand users.
Start free trial- Runs / month500
- Projects5
- SkillsUnlimited
- Ask, Show me, Do it
- Guardrails & session replay
- Remove Wizardful badge
- SupportEmail
Growth
Most popularThe copilot becomes part of how your product works.
Start free trial- Runs / month5,000
- Projects10
- SkillsUnlimited
- Ask, Show me, Do it
- Guardrails & session replay
- Remove Wizardful badge
- SupportPriority
Enterprise
Custom volume, custom terms.
Talk to us- Runs / monthCustom
- ProjectsCustom
- SkillsUnlimited
- Ask, Show me, Do it
- Guardrails & session replay
- Remove Wizardful badge
- SupportDedicated
- Every project gets a localhost-only test key. Testing never costs you a run.
- Run out of runs and the widget quietly disables itself. It will never break your page.
Questions
The ones that decide it.