BRG
The advisory intelligence machine for UK SME firms.
Visit →ATTOH · INTELLIGENCE
The AI operating layer underneath every ATTOH surface — Snap, BRG, CrewOS, Pharoah, Sentinel. One umbrella; five product lines; one decision per app.
ATTOH Tech is the build studio inside AMAYA Holdings. We build AI as outcomes — not seats, not API tokens, not empty chat boxes. Each app is scoped to one operator decision and ships with the source-of-truth, the supervision and the fallback baked in. The Snap suite is what is live today — the platforms below are what is coming next.
The advisory intelligence machine for UK SME firms.
Visit →The gamified AI incubator for ambitious operators.
Visit →Cinema-grade products for the Costa del Sol economy.
Visit →22 outcome-based apps across 14 UK SME sectors.
Visit →Counter-intelligence engine for high-value targets.
Brief us →
Most AI is priced like infrastructure: per seat, per token, per call. The buyer
operates the tool, prompts it well, supervises it, and absorbs the risk when it
gets the answer wrong. The bill arrives regardless.
ATTOH builds the other model. Every app on this page is scoped to one outcome —
a filing completed, a deadline cleared, a margin defended, a risk flagged before
it bites. The operator presses one button. The outcome is the deliverable. The
supervision, the source-freshness, and the fallback when one model fails — that
is our job, not the buyer's.
Every Snap is scoped to a single operator question — "Am I exposed under DSAR?", "What is my retention risk on this job?", "Is this tenancy RRA-compliant?". Nothing ships as a general-purpose assistant. The work is to narrow the surface until the answer is unambiguous.
The outcome only counts if it is right. Every app is wired to official UK endpoints — HMRC, Companies House, ICO, sector regulators — with weekly source- freshness CI. When a rule moves, the affected calculator is flagged and patched. No advice goes stale silently.
No empty chat boxes. No "tell me what you want". No infinite suggestions. Each app opens on the question, asks for the few inputs that matter, and returns a decision the operator can act on the same morning. Built for the day-to-day, not for casual exploration.
Not a transcript, not a prompt-engineered response, not a "draft for your review". The app does the work — the operator gets the decision, the document, the number or the flag. Human review sits at the escalation gate, where it earns its time.
Claude primary, multi-model fallback, shared governance routes, decision log, audit record. When a foundation model has a bad day, the app does not. Buyers should not have to think about prompt versions, context windows or rate limits to get a working answer.
We would rather ship 22 focused apps that each solve one decision than one monolithic platform that promises everything. Narrow scope means cleaner accuracy, easier QA, and a price tag tied to a result the operator can name.
Everything live today sits on one of three shelves. Snap apps answer a decision. Forge apps ship an artefact. Builders ship the brand surface around both. One engineering core underneath all three, audited once, reused across the suite.
Each Snap is a focused, free-to-try decision-support app for one industry — built on the same engineering core, hardened with backend governance, and designed to take an operator from question to outcome in minutes, not consultant hours.
Where the Snap line answers a question, the Forge line ships the document. Each Forge app takes an operator from blank page to a sendable artefact — a quote, a job spec, an invoice, a course outline, a client portal — in minutes. Built on the same engineering core, priced against the artefact produced.
Content drafting at scale — blog, email, social, briefs and sector explainers, voiced and structured for the audience that has to read them.
Proposals and project quotes that win — used downstream of TradeSnap, FreelanceSnap and HRSnap when the next step is a sendable document.
Job specifications and the recruitment paperwork around them — pairs with RecruitSnap and HRSnap when the operator needs the ad live the same day.
Invoice and billing artefacts — sector-aware line items, payment terms, and the chase sequence around them.
Client portal scaffolding — branded, secure, ready to drop on a domain. The asset operators and accountants ship to their own clients.
Course outlines and training material — modules, lesson plans, assessment briefs, sequenced for trainers and L&D leads.
Review and feedback collection flows — request, capture, route, publish. The artefact is verified social proof on the operator's site.
Where Snap and Forge are individual apps, these are full operating layers — productised for a vertical, a partner, or a buyer who needs the whole stack rebadged. Built on the same engineering core, shipped under their own brand and their own commercial route.
The gamified AI ops layer for the people who actually run the business day to day — leaderboards, competitive workflows, operator-grade dashboards. Built for the team floor, not the boardroom.
Close-protection AI ops platform. Enterprise tier, white-labelled to partners in the secure-operations sector. Public surface kept deliberately quiet — partner briefing under NDA on request.
The self-spec builder behind ATTOH Digital. Operators choose style, sector, features and price band — the configurator outputs a build brief our team can quote off. Live at attoh.digital/configurator.
Self-spec your next vertical app. Operator-driven brief generator that scopes a new Snap or Forge — sector, decision, sources, output — into a build outline the team can ship from. Multi-model proxy + Netlify deploy in commissioning.
UK talent and employment operating layer — placement workflows, contract pipelines, IR35 routing, candidate scoring. Pairs with RecruitSnap, HRSnap and HireForge as the end-to-end hiring surface for partner firms.
Vertical operating layer for gym and fitness operators — membership, scheduling, retention signals, instructor pipeline. The first of the vertical-OS line built on the engine. Repo scoped, build kicks off Q3 2026.
Same engineering core, different audience tier. Three brands sit on the build studio — from white-label partner bundles through premium hand-coded delivery — depending on who you are and how visibly we sign the work.
The flagship agency surface — custom-coded SME websites, AI Business-in-a-Box bundles, the configurator that lets operators self-spec their build. Premium tier, ATTOH-signed.
Hand-coded websites and sector app layers — the engineering surface that ships premium digital for agencies and partner firms who want our team behind the door but their brand on the front.
The white-label website service for partner firms — accountants, law firms and consultancies who sell websites to their own client base on the ATTOH build engine, under their own brand and pricing.
Snap apps, Forge apps and the Platforms all run on one shared engineering layer — an agent fleet, a multi-model routing gateway, an MCP server for AI-callable intelligence, a 49-strong CLI toolbox, and a Postgres brain with 480+ migrations underneath the lot. Built once, audited once, reused across the suite.
Each department owns one operator outcome category. Each agent inside it is narrow-scoped, supervised, dual-run-safe, and routed to the cheapest model that meets the quality bar. The architecture is public — the customer integrations underneath stay private.
Weekly sector synthesis · regulatory scans · competitive landscape · opportunity shortlists. The agents that read the world so the operator does not have to.
Tender · planning · distress-signal · Companies House mining → structured-API verification → autonomous LinkedIn outreach with split writer/dispatcher gates.
Seven agents covering company-filing deadlines, R&D eligibility, IR35 risk, insurance gaps, HMRC notices, regulatory drift, and grant scheme tracking.
Three-agent content cell — drafting, SEO & distribution, and quality gate before anything leaves the building. Voice and structure trained per audience.
Bill-of-quantity pricing with untrusted-document supplier-quote intake and scenario narrative — the build cost agent for ambitious construction operators.
Cash-forecasting agent reads ledger data, surfaces driver-not-restatement variance commentary, and flags emerging cash-low or critical-debtor positions.
Weekly pipeline-review agent — deal scoring, cross-sell detection, next-action recommendation. Internal pipeline data only, no external untrusted ingest.
Managed-agent leaf-worker orchestrator co-ordinates the cross-department briefs. Skills library is the source of truth; agent bundles are synced not copied.
Every agent task routes through a multi-model gateway. Synthesis goes to one model, matching to another, subagent untrusted-document reading to a third, and only the high-stakes outputs in regulated sectors reach a frontier model. The model layer is our problem — operators never see it.
Bulk weekly synthesis, long-context briefings, the workhorse for everything non-regulated. Strong reasoning at a fraction of frontier-model cost.
Scoring, ranking and opportunity matching — the agent muscle that compares one candidate against many and returns a ranked shortlist.
Fast, cheap subagent calls — untrusted-document reading with schema-validated JSON output only, no MCP access, no write scopes.
Reserved for regulated-sector outputs and customer-facing voice — medical, legal, finance review. The high-trust layer, used sparingly by design.
Vector ops for semantic search, deduping, narrative-cycle detection — the lowest cost tier in the stack.
If one foundation model has a bad day, the agent does not. Per-task fallback routes baked into the gateway; quota and budget enforced at config level.
The ATTOH MCP server exposes our sector intelligence to any Model Context Protocol-aware assistant. Eight tools, tiered access, rate-limited and security-hardened. Built on Express plus the official MCP SDK, deployed on Railway. Partner integration on request.
| check_company_compliance | filing & statutory health |
| get_tax_intelligence | sector-tuned tax posture |
| match_lenders | SME finance readiness scoring |
| find_grants | eligible grant scheme detection |
| check_ir35_risk | contractor risk by sector |
| get_insurance_gaps | mandatory cover gap analysis |
| get_local_seo_audit | local search visibility scan |
| get_sector_intelligence | cross-source sector brief |
| Protocol | Model Context Protocol (official SDK) |
| Runtime | Express on Railway |
| Access | Tiered subscription · API-key gated |
| Hardening | Rate-limited · security headers · audit log |
| Audience | Partner AI assistants & agent stacks |
| Integration brief | NDA-gated · email for spec |
Every operating capability has a CLI. Reusable from any machine via pip install, scriptable into n8n flows or shell pipelines, used internally and by white-label partners to drive content, outreach, intelligence and pipeline operations at scale.
AI spokesperson · voiceover · script generation · video scripts · render benchmarks · royalty-free clip harvester.
SEO blog · press releases · case studies · job ads · proposals · pitch decks · captions · content calendar.
SMS · WhatsApp · email drip · cold outbound · LinkedIn rewrite · multi-platform publishing.
Chatbot builder · landing pages · booking · funnel builder · Stripe billing · bolt-on shop · referral · reseller · client onboarding.
Prospect engine · competitor tracker · GA4 ROI · lead scoring · sector news · review responder · pain-point sync · ROI calc.
Performance brain · weekly digest · churn detection · recession pack · agency ops · portal · QC pipeline · lead injection · health checks.
One Postgres schema underneath everything — apps, agents, dashboards, hubs. Every agent reads from it, every operator action writes to it. The flywheel turns on one shared brain rather than seven disconnected ones.
| Runtime | Supabase Postgres |
| Migrations | 480+ applied, source-of-truth in git |
| Multi-tenancy | Row-level security, per-partner partition |
| Compound views | Cross-vertical intelligence aggregations |
| Source freshness | Weekly CI against official endpoints |
| Sandbox tables | Per-agent _sandbox parallel to prod |
| Dual-run views | Diff status per partner_id |
| Cutover gate | 3 aligned days + named sign-off |
| Drift detection | Nightly alert on status != aligned |
| Greenfield agents | Spot-check path when no prod twin |
These are the surfaces sitting on the same engineering core — agency sites, partner white-labels, sector hubs and the configurator. Each one is a live attestation of the stack: standalone HTML/CSS/JS on the front, FastAPI + Supabase + Cloudflare on the back, the agent fleet doing the silent work underneath.
Every Snap follows the same path from blank page to live app. Tight scopes, hardened backends, sector-accurate copy, governance baked in. No prototypes left rotting in staging.
Every app starts from one operator decision worth getting right — a deadline, a margin call, a compliance question. We work the scope down until the app does that one job better than anything else on the market.
FastAPI backend with hardened contracts, vanilla HTML/CSS/JS frontend (no framework lock-in), shared governance routes for audit, decision log, payment gates and human-approval gates. Every Snap ships with the same operating spine.
Every calculator is wired to a live source registry — HMRC, Companies House, ICO, sector regulators, official statutory instruments. When a rule changes, the affected app gets flagged and patched. No advice goes stale silently.
Static frontends on Cloudflare Pages. Python backends on managed infra. No build step, no bundler, no React rewrite waiting to happen. Each app is standalone — moves between domains, hosts, even owners, without a rewrite.
Each item below is one new outcome we are scoping into the suite — a decision an operator currently outsources to a consultant, a spreadsheet, or hope. Some are in private beta, some in design, some in commercial scoping. If one of these is your problem, talk to us early.
Long-form research surface for UK SME advisory — sector briefings, regulatory drift notes, "what changed" round-ups. Authoring tools and distribution layer in build.
The portal that ties the Snap suite, the AMAYA hubs, and partner integrations into one operator dashboard. Lead routing, decision log, and client surface in one place.
SME cyber posture and insurance comparison tools — sized for owner-managed businesses that are too small for enterprise vendors and too exposed for spreadsheets.
Auto-enrolment readiness, payroll error catch, pension contribution check — the small calls that cost owner-managed businesses real money when they slip.
Owner-managed tax and IP toolset — R&D credit eligibility, VAT registration triggers, trademark availability, immigration sponsor compliance. Four Snaps, one programme.
The Snap engine, rebadged. Accountants, law firms and consultancies can ship their own branded decision-support app on our spine. Scoped now, build in stages.
ATTOH Tech is the build studio inside AMAYA Holdings — a private UK operating group put together by Craig Attoh after years of watching SMEs buy AI by the seat, the token and the API call, and absorb every miss themselves.
The bet is simple. The current AI market sells infrastructure and asks the owner-managed business to operate it. That model leaks money — wrong prompts, missed updates, unsupervised drift — and the buyer carries the risk. We sell outcomes. Each Snap, each Forge, each platform on this page is scoped to one decision the operator can name, priced against the result it produces, and supervised end-to-end so the model is our problem rather than theirs.
That bet has shipped 22 Snap apps, a 7-app Forge line, six platforms in build, an agent fleet across 7 departments, and a shared Postgres brain with 480+ migrations underneath. The sister surfaces — attoh.digital, attoh.casa, attoh.eu — all sit on the same engine.
If you want a partner who builds AI as a result you can point at, not a tool you have to babysit, you are in the right place.
One shared stack, audited once, reused across every Snap. No bespoke framework per app, no consultancy lock-in, no rewrite waiting in the wings.
| Markup | Vanilla HTML5 · semantic |
| Styling | Hand-written CSS · design tokens |
| Interaction | Vanilla JS · no framework |
| Hosting | Cloudflare Pages · static |
| Accessibility | WCAG-aligned · automated smoke tests |
| Performance budget | No build step · no bundler tax |
| Runtime | Python 3.10+ · FastAPI |
| Governance contract | Shared routes · decision log · audit record |
| Database | Supabase Postgres · RLS |
| Source freshness | Weekly CI against official endpoints |
| Schedulers | APScheduler · cron · n8n |
| AI layer | Claude · multi-model fallback for resilience |
We sell results — not seats, tokens or chat boxes. Pick the route that matches the decision you need an app to make. Each tier is fixed-scope: you know what ships, when, and what it costs before we start.
Pick a Snap off the shelf. Free to try. Pay only when the answer ships.
A new Snap or Forge for your sector. Scoped, shipped, supervised — 4–8 weeks.
Run the ATTOH engine under your own brand. White-label app studio + agent fleet.
ATTOH Tech ships custom apps for UK SMEs. Sometimes that means picking a Snap off the shelf and configuring it for your business. Sometimes it means designing a new vertical app from scratch. Sometimes it means quietly white-labelling the engine for an accountancy or law firm that wants to own the decision surface in its sector.
30 minutes, plain English, no decks. Tell us the decision you need an app to make — we will tell you whether it is a Snap configure, a new build, or a roadmap item we can pull forward.
Display · Syne 700/800
Body · Inter 300-600
Mono · JetBrains Mono 400/500
Field notes from operators and partner firms running on the engine. Sector and surname held back where commercial sensitivity warrants it — full reference available under NDA.
"We stopped buying AI seats. ATTOH ships the answer — we pay against the result. The accountancy team gets a DSAR call cleared in minutes instead of hours and the bill maps to a thing we can name."
"FormSnap and AddressSnap dropped into our formations flow on a Friday and saved us a full FTE by the next quarter. The source registry is the killer — we never send a stale rule on a registration."
"I asked for a recruitment Snap. They came back with RecruitSnap, HireForge and a white-label site on the same engine — live in eight weeks. The supervision sits with them, not us, which is why the board signed it off."
"We are in a sector where wrong advice is a tribunal away. The freshness CI and audit log are why I sleep at night — the model is their problem, the decision is still ours, and we have the trail."
Quick answers to the questions we hear most often. Anything deeper, just email and we will get into it on a 30-minute call.
ATTOH Tech ships under the AMAYA Holdings group. Each surface below has its own audience, its own job, and its own commercial route.
ATTOH Tech ships AI as a result the operator can act on. One app, one decision, one answer — sourced to the day it ships, supervised end-to-end, priced against the outcome it produces.