AI-assisted delivery,
with a record your clients can read.
For software agencies (10–500 developers) where Claude, Cursor, Copilot, and Codeium are now writing meaningful chunks of customer code. When your client asks "show me what your AI did on our project — and who signed off," Xlooop has the answer ready: every prompt, every change, every reviewer, captured as work happens. No retracing. No audit gap. No liability surprise at handover. Built on our bi-directional code↔UI engine — on top of a patented multi-user, role-based collaboration model (US 11,604,641) — detailed below. AWS later named this lifecycle AI-DLC — and our role-based collaboration model has been patented since 2018 (US 11,604,641), years before the category had a name.
discovery · A$8–15k · pilot · 4–6 weeks · sterile teardown if the fit isn't there
Software development is where the Xlooop engine was forged.
This is the discipline where the bi-directional code↔UI engine and the multi-agent delivery model were forged. Across 130+ enterprise engagements at our Sydney and Melbourne consulting practice (2011–present), we made the mistakes, learned the patterns, and shipped the result. The bi-directional engine and multi-agent model were proven here first — today that same foundation runs across our regulated-SMB verticals (accounting, construction, advisory), with software-development one vertical we partner on selectively at early stage.
Xlooop is one product applied across multiple verticals — software-development (this page), accounting, construction, and advisory. The technical foundation (bi-directional engine architecture, audit-grade evidence trail, 6-role panel) is shared. Prompts instead of customer letters. Diffs instead of workpapers. Storybook snapshots instead of disclosure packs. Only the lexicon and the surface change.
Other vertical surfaces →Xlooop overview·accounting·construction
Our bi-directional engine, built on a patented collaboration model.
Xlooop is not an AI wrapper. The product is built on a 3-engine architecture (~12 months of in-house engineering, on top of 14+ years of enterprise delivery patterns at our Sydney + Melbourne consulting practice). Xlooop owns a granted US patent and a European pending patent covering its multi-user, role-based collaborative model and feedback-trained engine with chat-bot integration (US 11,604,641, Claim 1; 24 claims, 2 independent; granted 2023) — also covering generation of the UI in deployment-ready code (Claim 23). The bi-directional code↔visual sync is our own engine architecture, built on top of that patented foundation. Every claim on this page traces back to engines that exist in our codebase today.
Three layers: a UI layer with a visual editor and multi-user roles — the multi-user role model is the subject of US Patent 11,604,641, Claim 1 — an AST-proxy layer with compiler, editor and codabase engines, and a code layer with the source repository and an audit-grade evidence ledger. Bidirectional arrows show that visual edits write back to source code without breaking the AST.
Hover or tap any engine node
Visual edits write back to source code through our AST proxy engine — no design-vs-code drift. The animation walks Capture → Govern → Sign.
CompilerEngine
Source-fidelity round-trip: JSX/TS → AST proxy → renderable artefact. The other engines and the agents all operate on the same ground-truth AST — no translation loss, no "design vs code" drift.
EditorEngine
Visual editing with write-back through a mutation dispatcher — your designers, PMs, and clients can adjust the visual representation; changes propagate back to source code without breaking the AST.
CodabaseEngine
Codebase-graph-aware AST manipulation. Refactor-safe changes across multiple files, architectural-pattern awareness, multi-project continuity — the foundation for delivery without breaking things at scale.
Our bi-directional sync, built on a patented collaboration model
Cursor, Copilot, Claude, Devin generate code. We don't replace them. We capture what they produce, govern it through our bi-directional engines, and preserve it as audit-grade evidence — across every project your agency runs in parallel, across every team rotation, across every client review cycle. Real-time bi-directional sync between the code layer and the visual/role-based editing layer is the architecture we built on top of our patented foundation. What's patented is the multi-user, role-based collaborative model and feedback-trained engine with chat-bot integration — US Patent 11,604,641, Claim 1 (granted March 14, 2023; 24 claims; 2 independent claims), which also covers generating the UI in deployment-ready code (Claim 23) — plus European patent pending. Further patent details available under NDA via the investor data room.
A role-governed delivery model — patented before AWS named the category.
AWS later formalised the AI-Driven Development Lifecycle as AI-DLC — AI execution with mandatory human oversight, decisions deferred to people, and persistent context across every phase. Our patent (US 11,604,641, priority 2018, granted 2023) covers the multi-user, role-based collaboration model and feedback-trained engine — which built exactly those human-oversight controls into AI-assisted delivery years before the category had a name. (The three bi-directional engines above are our own architecture, built on top of that patented foundation — not themselves the patent.) AI-DLC names the lifecycle; XCP (Execution Control Plane) is how delivery is governed and proven — a risk-based, graduated governance model, an enforced governance cadence, and a closed-loop evidence engine. Full rubric under NDA.
One persistent system of record across every client engagement.
Knowledge stops being scattered across tickets, documents, calls, and individuals' heads. Every requirement, every design decision, every implementation choice — linked + preserved — so continuity survives team rotation, client change-requests, and multi-project parallelism.
Onboarding new developers
A new joiner doesn't interview the existing team to learn "why this decision was made." The full context is in Xlooop — captured at the moment the decision happened.
Team rotation
An outgoing developer's context doesn't leave when they do. Codebase-graph-aware capture means handoffs are smooth, not catastrophic.
Customer change-requests
When a client asks "remind me what we agreed on this?" the answer is one query away, with evidence — not a Slack archeology expedition.
Margin protection
Reducing rework rate is the single biggest commercial driver — every agency lead we’ve worked with treats it as their margin lever. Bi-directional sync eliminates design-to-code translation overhead at the architecture layer, not the tool layer.
Multi-project parallel delivery
A single system of record across all active engagements. No context-switch overhead between projects. No "wait, which client was that for?" moments.
Client trust at sign-off
Every AI-assisted decision is signable, hash-chained, and exportable. Your client gets the audit pack; you keep the codebase ownership; the IP belongs to neither us nor them.
A real feature delivery, captured and governed — live.
These widgets are a live render of the product surface. The activity stream and role panel below mirror what your team would see during an actual Xlooop-governed delivery. No screenshots. Animated from real data types.
Every prompt, diff, and reviewer action is captured to a write-only evidence ledger — role-attributed and hash-chained. Your client sees exactly what AI did and what a human signed off on.
Each AI agent is bound to its role: the Ecosystem Risk Officer flags exposure, DevSecOps owns commits and scopes, the Commercial Claim Reviewer holds the final veto — no single agent can approve its own work. The Delivery Gate only opens when evidence is complete.
Four problems we hear, every conversation.
Velocity without traceability
Your delivery team uses Claude, Cursor, Copilot, Codeium — each spinning up its own context. Six weeks in, no one can answer "which prompt produced this function" without re-reading every PR.
Review debt
AI-generated PRs are bigger and faster than human ones. Reviewers either rubber-stamp or burn out. The cost shows up later as production drift, hotfix cycles, and brittle dependencies.
Client trust questions
"How much of this codebase did your AI write?" is now an SOW conversation. Without a clean record, you either over-disclose or under-deliver — both kill the renewal.
Disclosure language
When the client's legal team asks for an AI-usage attestation, you need a numbered, signable, dated answer — not a screenshot of your Linear board.
Capture · Govern · Sign.
The same three beats your delivery process already has — only now, audit-graded.
Capture
Every prompt, diff, and review event from your delivery team is captured in a sterile, per-tenant ledger. No model-training, no harvest, no external mirror.
Govern
Workflow agents review for unsupported claims, license issues, and dependency drift. Human reviewers see a pre-filtered diff with citations.
Sign
A signable evidence pack — per feature, per release, per client — produced on demand. Hash-chained. Region-bound. Yours, not ours.
Capture · Govern · Sign is how XCP governs the lifecycle AWS later named AI-DLC: AI execution with mandatory human oversight, decisions deferred to people, and persistent context across every phase. Our role-based collaboration model has been patented since 2018 (US 11,604,641) — years before the category had a name. AI-DLC names the lifecycle; XCP is how it is governed and proven — risk-based graduated gates, governance cadence, closed-loop evidence engine. Full rubric under NDA.
Four specialists, tuned to delivery work.
Distinct from the Phase-A welcome panel — these agents are the ones an agency lead actually meets in week one.
Delivery Auditor
Captures every AI-assisted prompt → diff → test run → reviewer signature into a single per-feature ledger.
Review Companion
Surfaces unsupported claims, dependency drift, license issues, and missing test coverage before the human reviewer opens the PR.
Storybook Reconciler
Keeps component stories aligned with the generated code; flags drift, regenerates snapshots, lists visual diffs for sign-off.
Disclosure Drafter
Produces the AI-usage attestation your client legal team asks for — per project, per quarter, per regulator if relevant.
One discovery. One pilot. Two go/no-go decisions.
We don't ship sandboxes. The discovery and pilot run on a real tenant of yours — sterile teardown returns the work if you decide not to continue.
Discovery
We walk one of your live client engagements end-to-end with your delivery lead and reviewers. Output: a written diagnostic, a workflow map, and a written pilot proposal — or a written "no fit, here's why".
Pilot
A single feature stream — one team, one client — instrumented with Xlooop. We turn on Capture in week 1, Govern in week 2, Sign in week 4. Pilot ends with a signed evidence pack for the client.
Sterile teardown
If the pilot doesn't earn its keep, your tenant is wound down within 14 days. Export bundle produced. Hashes verified. Proof-of-deletion certificate at the 30-day mark.
pricing in AUD · invoiced net 14 · we do not require a pre-paid annual licence before pilot graduation.
The shape of what we don't do.
- NOT CURSOR / COPILOT WORKSPACESwe don't replace your IDE. We sit alongside it and govern the work it produces.
- NOT A DELIVERY FRAMEWORKwe don't prescribe Scrum, Shape Up, or anything else. Bring your process; we instrument it.
- NOT A "RIP-AND-REPLACE" CIwe run beside GitHub Actions / GitLab CI / Buildkite. We don't take over your build.
- NOT A MODEL HARNESSwe never train external models on your client's code. Not now, not later, not as a default.
- NOT A STAFFING POOLwe don't place engineers. Your team stays your team — we add an audit-grade ledger on top.
One conversation. One client engagement. One pilot.
We partner selectively with software-development teams where the AI-agentic layer materially compresses delivery time, raises code quality, and produces a clean audit trail of every AI-assisted decision. At early stage, we maintain a deliberately small cohort — protecting both the operating team's focus and the IP integrity of the technology stack.
Deliberately small early-stage cohort · operator-led intake · IP-integrity-protected