Agent demo lab
Live AI agent demos. Don't take our word for it.
Put all 16 of our AI agents to work for fictional businesses: call a clinic, fix a damaged order, book a site visit, write a week of posts, move an ad budget, process invoices, query a database, connect tools over MCP, or try to jailbreak a bank's chatbot. Each one uses real tools, follows real rules and shows you every step it takes and why.
- Real AI, live, no sign-up
- Guardrails and approval gates you can test
- A trace of every tool call
Live now
Pick an agent and put it to work.
Answer · Every call, chat and message, handled.
Receptionist Agent
Harbourside Dental · fictional
Talk to a real AI voice receptionist for a fictional dental clinic.
You play a patient calling the clinic before it opens.
Try it liveSupport Agent
Kettle & Pine · fictional
Chat with an AI support agent for a fictional online store.
You play a customer of an online homeware store.
Try it liveWhatsApp Agent
Oud & Linen · fictional
Chat with a WhatsApp-style AI agent for a fictional Dubai boutique, in English, Arabic or Urdu.
You play a shopper who left a kaftan in the basket yesterday.
Try it live
Sell · Speed-to-lead and booked meetings.
Booking Agent
Cedar & Stone Renovations · fictional
Send a sample enquiry to a fictional remodeling firm and the AI booking agent texts you back within seconds.
You play a homeowner who just sent an enquiry from the website.
Try it liveSDR Agent
Northbeam · fictional
Pick a sample company and the AI SDR agent researches it, checks the fit, finds the right person (respecting a do-not-contact list) and drafts a short, specific email that must cite a real buying signal.
You play a sales rep at a B2B software company.
Try it live
Market · Content, search, ads and reviews.
Social Media Agent
Basil & Brick · fictional
Brief the AI social media agent for a fictional Manchester pizzeria and it plans a week of Instagram and X posts in the brand voice, checks each one against platform limits and allergy rules, and answers comments.
You play the owner of a local pizzeria.
Try it liveSEO & GEO Agent
Atlas Bike Repair · fictional
The AI SEO & GEO agent audits a fictional bike shop’s pages in code, shows keyword gaps and whether AI assistants cite the business, then drafts titles, descriptions and FAQs that are validated before you approve publishing.
You play the owner of a local bike repair shop.
Try it liveAds Agent
Peak Trail Outfitters · fictional
The AI ads agent reads a sample Meta and Google Ads account, explains ROAS per campaign and proposes budget moves from losers to winners.
You play the marketing lead of an online store.
Try it liveReputation Agent
Juniper Café · fictional
The AI reputation agent drafts on-brand replies to a fictional café’s Google reviews, posts the positive ones, sends critical ones to the owner for approval, reports fake or spam reviews and asks every recent customer for a review, the way Google’s rules require.
You play the café owner.
Try it live
Operate · Documents, data and back-office work.
Knowledge Agent
Fernhill Logistics · fictional
Ask a fictional company’s policies anything: leave, expenses, travel, IT security.
You play an employee (or switch to manager).
Try it liveDocument Agent
Fernhill Logistics · fictional
The AI document agent reads sample supplier invoices, extracts the fields into structured JSON and validates them in code: line maths, VAT, totals, purchase-order match and supplier.
You play an accounts payable clerk.
Try it liveOps Agent
Brightline Print Co. · fictional
The AI ops agent runs a fictional print shop’s order-to-invoice workflow: it reads the email, matches the customer in the CRM, prices the job from the price list, creates the invoice, logs it, notifies the team and asks you before emailing the customer.
You play the operations manager.
Try it liveData Analyst Agent
Roastery Lane · fictional
Ask a fictional coffee shop’s database anything: revenue by month, best sellers, customer countries.
You play the founder asking business questions.
Try it live
Secure and Build · Connect, harden and ship AI products.
MCP Connector
Orbit Labs · fictional
An AI assistant answers questions by calling a fictional company’s CRM, support tickets and docs through MCP servers.
You play an account manager using an internal AI assistant.
Try it liveGuardrail & Eval Audit
Nimbus Bank · fictional
A fictional bank’s AI assistant hides a secret code.
You play an attacker trying to break a protected chatbot.
Try it liveAI MVP Sprint
Visionbotix MVP scoping
Describe your AI product idea and our scoping agent interviews you, cuts it to a launchable MVP of five core features, lays out the build plan and prices it with the same estimator our team uses.
You play a founder with an AI product idea.
Try it live
Site concierge
Visionbotix · this website
Our own AI concierge answers questions about our agents, prices and process from this website, and sets up a call. It's in the corner of every page.
How the lab works
Real agents, safe sandbox.
Every demo runs on the same platform we use to build agents for clients: provider-agnostic, tool-based, with guardrails in code and a full audit trail.
- 01
Checked before the model sees it
A one-time Cloudflare human check, rate limits, a length cap, card numbers and keys removed, and Llama Prompt Guard scoring the message for prompt injection.
- 02
The agent picks its tools
A large language model reads the conversation and decides which of its allow-listed tools to call: calendar, orders, catalog, CRM or documents. It must give a reason for each call.
- 03
Rules run in code
Each tool checks its own rules (identity, limits, stock, access rights) on fictional sample data. Anything over a limit stops at an approval gate for a person.
- 04
Checked on the way out
The reply is screened for leaked instructions before you see it, and every step lands in the trace beside the conversation.
FAQ
About the demos.
Are these demos real AI agents?
Yes. Each reply comes live from a large language model (by default the open-weight gpt-oss-120b, served by Groq or Cerebras, with Google Gemini as a backup) that chooses and calls real tools. Only the businesses and their data are fictional, so nothing you do has consequences.
Which model do your production agents use?
The one that fits the job and your data rules: Claude, GPT, Gemini, open-weight models such as gpt-oss or Llama, or a model running on your own servers. Our agent platform is provider-agnostic, so switching is a configuration change.
What stops an agent from doing something it should not?
Four layers: checks on every incoming message (including a prompt-injection classifier), tools that enforce their own rules in code, approval gates for risky actions, and a check on every reply. Every step is logged in an audit trail, which the demos show as the trace.
What happens to what I type in a demo?
It is processed to answer you and kept briefly in our host’s logs to stop abuse and improve the demos, never used for training. Use made-up details. Details are in the privacy policy.
Is there a demo for every agent?
Yes. All 16 packaged agents have a live demo, from the voice receptionist to the MCP connector and the prompt-injection challenge, plus the site concierge in the corner of every page.
Seen enough? Let's build yours.
Book a free 30-minute call. We'll pick the right agent for your business and quote a fixed price within 24 hours.
NDA on request from the first call · You own 100% of the code and IP