01 SEPT 2026 ·AI
What is WebMCP? Why it matters for websites and AI agents
A practical guide to WebMCP, how it differs from MCP and browser automation, and what we learned connecting Chat Thing to page-declared tools.
Jozef
AI agent development · UK
AI agent development for UK businesses
AI agents that ship.
An AI agent development company in Newquay, Cornwall, building custom AI agents for UK businesses. Trained on your data, wired into your systems and built to do real work.
Talk to us What you get
Everything an agent needs to actually work.
A real agent needs more than a model and a prompt. It needs tools it can call, content it can trust, evaluations that catch regressions, and a team that maintains it. Here's what we deliver.
01
A production-ready agent
A fully built agent, deployed to your stack with monitoring, evaluations, and a clear hand-off, and ready to handle real traffic from day one.
02
Tools that match your workflows
Agents only work when they can actually do things. We connect them to your APIs, data sources, and SaaS tools (CRM, ticketing, billing, internal services) so they can take action as well as answer questions.
03
A controllable knowledge layer
Retrieval-augmented generation tuned to your content, your tone, and your edge cases. We set up the ingestion pipeline so your agent stays current as your docs and data change.
04
Evaluation and guardrails
Automated evaluations so you know whether changes make the agent better or worse. Hard guardrails on the things that matter: escalation rules, refusal cases, audit trails.
05
A dashboard that tells the truth
Conversations, tool calls, costs, latency, success metrics. So you can see what your agent is doing in production and where it needs improvement.
06
A team that stays around
Agents need looking after once they are live. We support and tune them after launch, with pricing set up for an ongoing relationship rather than a one-off project.
How we work
6 week product cycles that always launch.
Build your vision with our 6-week product cycles. A small senior team, AI-amplified end-to-end, geared up to launch your idea in six weeks.
Why 6 Weeks? It's the Goldilocks Zone - Striking the perfect balance between allowing enough time to build something meaningful, while being short enough to keep risks low!
Whether its an MVP, prototype, or feature in a existing product, our 6 week cycles make sure you have something tangible at the end of the project.
Sounds cool! Tell me more01: Discovery
Refine your ideas and plan what will be launched in 6 weeks.
02: Kick-off
We get cracking. Design, code, and AI work happen in parallel from day one.
03: Check-in
On week 3 get ready for an exciting demo of progress.
04: Build & Iterate
Continue work and integrate feedback from the check-in.
05: Pre-launch
A check-in before launch to tie up loose ends and get ready.
06: Launch
The big day is here, you idea is launched to the whole world.
Use cases
AI agents for business: what we build.
- 01
Customer support agents
Cut first-response time to seconds and only escalate the conversations that actually need a human. Trained on your help docs, ticket history, and product behaviour so answers stay accurate and on-brand. Our own Chat Thing runs support agents for businesses every day, and what we learn there goes into client builds.
- 02
Internal knowledge agents
Your team's instant answer to 'where's the doc on…?'. It searches what you've written down across Notion, Confluence, Drive, and Slack, and links back to the source so nobody has to take its word for it.
- 03
Slack, Discord & Telegram agents
Agents that live where your customers, community, or team already do. The same agent runs inside the chat platform, so nobody needs a new tab or another login.
- 04
Voice agents
Real-time voice agents for inbound support, outbound calls, and discovery. Fluent enough that callers don't realise they're talking to one. Connected to your CRM, calendar, and ticketing.
- 05
Embedded product copilots
The AI feature that's already on your roadmap, designed and shipped inside your product instead of bolted onto the side. We handle prompt design, retrieval, evaluation, and the UX so it feels native. If the agent lives inside your existing product, this overlaps with our AI integration services.
- +
Got something different?
Tell us about your use case. We'll come back with a straight answer about whether it's something we can help build.
Recent work
All work Real agents, in production.
We build our own AI agents alongside client engagements. Every lesson here gets folded back into the work we do for you.
DesignAIDevelopment
Chat Thing
Chat Thing is our first in house product that allows users to create magic AI chatbots
AI AutomationsAI product developmentAI
Building HeyYou a production AI security platform in 8 weeks
HeyYou had a working AI prototype. Pixelhop built the launchable product: a multi-tenant platform with deep Axis camera integration, all live in 8 weeks.
AI AutomationsAI agents
Orbit: AI grant scraping for Grants Online
An AI grant scraping pipeline that turned a year of manual review into minutes per grant.
Writing
All writing What we've learned building it.
Jozef
07 JUL 2026 ·Writing
Pixelhooks: a programmable endpoint for building with agents
Pixelhooks records every HTTP request sent to a throwaway URL and lets you program exactly what it sends back - a see-both-sides scratch endpoint for building against webhooks and agents, with an MCP server so your assistant can drive it.
Chris
12 JUN 2026 ·AI
Our first real test of Fable 5: brilliant, autonomous, and maybe too pricey to repeat
Our first real-world test of Anthropic's Fable 5: running inside Claude Code, it autonomously rebuilt a messy, already-deployed prototype onto our production stack in a single afternoon. Hugely impressive and near-faultless, but slow and notionally pricey, which raises a real question once API pricing lands.
Jozef Pricing
What it costs.
Most builds run in six-week cycles at £27,000 + VAT. That's a small senior team designing, building and launching one clear outcome. Most agents fit in one cycle; bigger products run over two or three.
Smaller one-off jobs are quoted individually at our day rate of £650 + VAT, even if it's only a couple of days' work.
Not sure what you need yet? We start with a discovery week (£3,250 + VAT) and finish it with a fixed-price proposal for the build.
After launch, many of our clients stay with us on a monthly retainer for support and improvements. Model, API and hosting costs are separate, and you can see them on the cost dashboard from day one.
FAQs
AI agent development FAQs.
How long does it take to build an AI agent?
A meaningful prototype runs in 1–2 weeks. A production-ready agent typically takes 4–8 weeks from kickoff, depending on how many tools it needs to call and how strict the evaluation bar is.
Which models and platforms do you use?
We're model-agnostic and work with Claude, GPT, Gemini, and open-source models depending on the job. Same for orchestration: Vercel AI SDK, custom stacks, the Anthropic Agent SDK, MCP servers, whatever fits.
What's the difference between an agent and a chatbot?
A chatbot replies. An agent acts. Agents have tools they can call (your APIs, your databases, third-party services) so they can do real work. Book a meeting, refund an order, escalate to a human, update a record. We build the second kind.
Where can I deploy my agent?
Anywhere your users already are. We've shipped agents into web chat widgets, Slack, Discord, Telegram, WhatsApp, real-time voice calls, and embedded inside products as native features. The brain stays the same; we wire it up to whichever surfaces fit your workflow.
Do you help write the prompts?
Yes, and we keep working on them after launch. We write the initial prompts, set up an evaluation harness so prompt changes can be measured against real conversations, and tune them as the agent meets real users. Most engagements include a regular prompt review post-launch.
What's context engineering and why does it matter?
Context engineering is the work of getting the right information in front of the model at the right moment. Your docs, your tool descriptions, the user's recent activity, the screen they're looking at. It's usually 80% of the difference between an agent that feels useful and one that feels generic. We design retrieval pipelines, tool schemas, and prompt structures so the relevant context is always there without overwhelming the model.
When shouldn't we build an agent?
If the job is a fixed sequence of steps with no judgement involved, a plain AI automation is cheaper and more reliable. Agents earn their keep when the work needs decisions, tools and back-and-forth. We'll tell you which one you need.
How much does it cost?
Most projects fit in one six-week cycle at £27,000 + VAT. Smaller jobs are quoted at our day rate of £650 + VAT, and a discovery week is £3,250 + VAT. See What it costs for the detail.
Do you work with businesses outside Cornwall?
Yes. We're based in Newquay, and most of our clients are elsewhere in the UK. We work remotely, with the kick-off, week-three check-in and launch on video calls.
How do you handle our data?
We're UK GDPR compliant. At the start of a project we agree what personal data the work needs, where it's stored and who can access it, and we build to that.
Now booking
Got a job for a robot?
Tell us what you're trying to ship. We'll come back with a straight answer.
Start a conversation