For homeowners, renters & their insurers

One phone scan. An insurer-ready contents inventory.

RakuAI Capture turns a 60–90 second phone scan of a room into an itemized, queryable contents inventory — with approximate measurements and estimated replacement values. An adjuster’s AI assistant can query the inventory over the MCP surface without a site visit. Same 3D Gaussian-splat reconstruction and MCP core as the rest of RakuAI — no new engine. Built for three moments: filing a claim, planning a move, and downsizing.

Phase 1 / pilot readiness — honest about what ships today versus what’s on the roadmap. We document a space and produce a structured inventory; we are not an insurer and do not set policy terms or settle claims.

Member of

Microsoft for Startups Read more ↓

Request a pilot → Try the capture flow →

From a phone in your hand to a list you can submit.

The same Raku Capture pipeline, one step at a time. Use the arrows, dots, or your keyboard.

Step 1

Scan the room

Walk a slow loop with your phone — about 60–90 seconds, guided on-screen. No tripod, no $300 rig, no special camera.

Step 2

Reconstruct

Frames and motion are reconstructed in the cloud into a 3D Gaussian splat — an explorable model of the actual space.

Step 3

Itemize with any LLM

An LLM walks the scene over the MCP surface and lists the visible objects with approximate measurements. Claude, ChatGPT, Gemini, or Copilot — your choice.

Step 4

Review & export

You edit the list, confirm values, and export a clean inventory to hand to your carrier. Every value is labelled an estimate — you stay in control.

Three moments people need this.

We deliberately ceded real-estate listing capture to the incumbents (Matterport and friends). This is the consumer-facing wedge nobody owns: a self-serve scan that produces a structured contents inventory for the moments an inventory actually matters.

Claims documentation

After a loss — fire, theft, water, storm — reconstructing what you owned from memory is painful and disputes are common. A scan taken at policy onboarding or renewal gives a timestamped, itemized record before anything happens. When a claim is filed, the adjuster’s AI assistant queries the stored splat over MCP: “List the furniture in this room with dimensions and estimated values.” Filing becomes “submit the inventory” instead of “remember everything you lost” — and the adjuster may never need to visit.

Moves

Movers, valuation coverage, and transit insurance all want a contents list. One scan per room produces an itemized manifest with rough dimensions — useful for movers' quotes and for documenting condition before and after the truck.

Downsizing

Clearing out a home — an estate, a parent moving into assisted living, a kid leaving for college — means deciding what to keep, sell, donate, or insure. A scan gives the whole family one shared, itemized view of what's in the room.

How one scan becomes an inventory.

It's the same Raku Capture pipeline used across the rest of the product — scan, reconstruct, query. Insurance is a wrapper on output, not a new engine.

Scan the room

Walk a slow loop with your phone, 60–90 seconds, guided on-screen. No special hardware.

Reconstruct

Frames + motion are reconstructed in the cloud into a 3D Gaussian splat — an explorable model of the space.

Itemize over MCP

An LLM (Claude, ChatGPT, Gemini, or Copilot) connects to the 17-native-tool MCP surface and walks the reconstructed room: listing visible objects, approximate dimensions, and estimated replacement values — without a site visit.

Review & export

You edit the list, confirm values, and export a clean inventory you can hand to your carrier. Every value is labelled an estimate.

What the output looks like.

A future-state illustration of a priced single-room inventory — what the document will show once a licensed replacement-cost source is wired. The dollar figures below are illustrative placeholders, not values RakuAI produces today. The real export the pipeline generates right now is itemized but unpriced — see “The real claim export” just below. Either way you review and edit the list before it goes anywhere.

Item	Approx. size	Est. replacement
3-seat sofa	210 × 90 cm	$1,200
55″ television	123 × 71 cm	$650
Coffee table	110 × 60 cm	$240
Bookshelf (filled)	80 × 200 cm	$380
Area rug	200 × 290 cm	$320
Estimated room total	—	$2,790

Every figure above is an estimate, not an appraisal — and, today, an illustrative placeholder rather than a generated value (pricing is gated, below). Object recognition and value estimation are imperfect — you review and correct the list before it goes anywhere. RakuAI does not set your coverage, determine claim payouts, or guarantee replacement cost; your insurer does.

The real claim export.

This is the actual document an adjuster’s AI assistant receives over MCP today — GET /api/v1/capture/{id}/claim/export, generated from a reconstructed scan. It itemizes contents with approximate dimensions and a transparent confidence. It is honest about what is not there yet: no dollar values, because no licensed replacement-cost source is wired. Nothing here is faked — an unpriced line is shown as “not priced,” never as $0.

Item	Approx. size	Confidence	Est. replacement
3-seat sofa	210 × 90 × 90 cm	0.80	Not priced
55″ television	123 × 71 × 8 cm	0.80	Not priced
Coffee table	110 × 45 × 60 cm	0.80	Not priced
Bookshelf	80 × 200 × 30 cm	0.80	Not priced
Area rug	200 × 1 × 290 cm	0.80	Not priced
Room total — 5 items	—	0.80 avg	Not priced

Confidence is a data-presence score, not a learned probability: 0.80 means the object carried both a declared type and usable dimensions from the reconstruction — never 1.0, which would imply an appraisal-grade identification we do not perform. Dimensions are converted from the reconstructed scene’s world units (approximate, not measured). The same document is available as carrier-ingestible JSON or CSV. The replacement-value column stays “not priced” until a licensed cost source lands — that is the one named, honest gate, and we will not fabricate a dollar figure to fill it.

The carrier-pays model, plainly.

The consumer scans for free at onboarding or renewal. The carrier pays per scan. The carrier’s return is lower loss-adjustment expense (LAE) and a shorter FNOL-to-payout cycle.

Carrier pays $5–$20 per scan

The consumer scans at onboarding or renewal — no charge to the policyholder. The carrier funds the scan as an acquisition and retention investment.

LAE reduction

Fewer adjuster touchpoints on policies where a pre-loss contents inventory already exists. Target pilot metric: ≥10% reduction in adjuster touchpoints on scanned policies versus a control group.

↓

Shorter FNOL-to-payout cycle

When a claim is filed, the adjuster’s AI queries the stored splat over MCP and gets a structured inventory in seconds — not days waiting for an in-person visit and manual writeup.

Backend cost per scan is estimated at $0.50–$2.00 (insurance scans are typically single-room, low frame count). No signed carrier partnership yet; this is the value model we are taking into our Phase-1 pilot conversations.

Why RakuAI for this, specifically.

The capture market is crowded, but the consumer self-serve scan → structured inventory pipeline is wide open.

Approach	Strength	Gap for contents inventory
Phone-photo list (DIY / email)	Free, everyone can do it	No measurements, no structure, easy to lose, painful to reconstruct after a loss
Adjuster visit	Authoritative	Loaded cost roughly $200–$500 per visit; only happens after a claim is filed
Matterport-style 3D walkthrough	High-fidelity capture	Built for real-estate listings — no itemized contents output, no carrier-friendly export
RakuAI Capture	Self-serve phone scan, splat fidelity, LLM-driven itemization over MCP	Honest framing: values are estimates, and the formal carrier-export format is still being built

What's real today, and what isn't.

We would rather under-promise. Here is the current line between shipping and roadmap.

Shipping today

Phone-based capture and cloud splat reconstruction (the Raku Capture flow).
Scene query over the MCP surface — an LLM can walk the reconstructed room and describe what's in it.
Multi-LLM support: Claude, ChatGPT, Gemini, and Copilot can all drive the runtime.

In flight / roadmap

A polished, one-page insurer-ready export (room thumbnail + itemized table + totals + scan metadata). The underlying data exists today; the formatted document is being built.
Carrier-specific inventory templates — handled as per-pilot work, since there is no single industry-standard format.
Per-scan metering and a consumer change-tracking tier (re-scan annually, see a diff).

What we don't have yet

SOC 2 certification. Large carriers require it before procurement; we run early pilots with InsurTechs while that readiness work proceeds in parallel.
Any signed carrier partnership. We do not claim endorsement by any insurer.
Appraisal authority. We document and estimate; we do not value or settle.

How the adjuster-assistant workflow fits in.

The MCP surface that powers Raku Capture is also how an adjuster’s AI assistant queries the scan. Here’s what that looks like in practice — and what’s honest about where it stands today.

Policyholder scans at onboarding

The consumer scans their room with the Raku Capture flow. The carrier embeds this as a free step in new-policy onboarding or annual renewal.

Splat stored on RakuAI cloud

The reconstructed 3D Gaussian splat is stored on RakuAI’s cloud with the policyholder’s consent. Raw frames are discarded after reconstruction by default. Scan-specific data-handling disclosures are being added to the privacy policy before any consumer scan goes live.

Adjuster’s AI queries over MCP

When a claim is filed, the adjuster’s AI assistant connects to RakuAI’s 17-native-tool MCP surface and walks the reconstructed room. A natural-language query — “List the contents of this room with approximate dimensions and estimated replacement values” — returns a structured inventory without a site visit. The adjuster reviews and edits rather than measuring from scratch.

Adjuster reviews and closes

The adjuster edits and confirms the list against the claim. Every value is labelled an estimate. RakuAI does not settle claims or determine payouts — that stays with the insurer.

Honest status of this workflow

Phase 1 / pilot readiness (current): The MCP query surface works today — an LLM can walk the reconstructed room and return a structured inventory. The carrier portal, per-scan metering, and formal insurer-ready export (one-page PDF with scan metadata) are in flight for the first pilot.
Not yet: SOC 2 certification (required by large carriers; early pilots target InsurTechs first). No signed carrier partnership as of today.
Your data: Scans are never shared with any third party without explicit consent. The carrier sees only the inventory document, not the raw scan, via a carrier review portal — until SOC 2 lands, that is the data-handling model.

For carriers and InsurTechs.

Three concrete ways to engage. The carrier model: carrier pays per scan ($5–$20), policyholder scans free at onboarding or renewal, carrier wins with lower LAE and faster FNOL-to-payout cycle. Phase 1 / pilot readiness — no signed carrier partnership yet.

1. Capability inquiry

You're evaluating a contents-capture layer and want a written response on how RakuAI maps to your product, your preferred inventory format, and your onboarding flow. Tell us your constraints; we'll reply with what's real today and what we'd build for a pilot.

2. Free 90-day pilot

One partner per quarter. We integrate the scan flow into your onboarding (white-label or sidecar web app), collect roughly 100 customer scans, and co-publish a case study. We don't charge for the pilot; we ask for UX-team access and permission to publish.

Discuss a pilot →

Early pilots run on RakuAI cloud with a carrier review portal until SOC 2 lands.

3. InsurTech integration

You already own claims orchestration, aerial/exterior data, or auto damage assessment — and have no interior contents story. We can be the capture-and-itemize layer you resell under your brand via API.

Explore an integration →

Request a free 90-day pilot.

One partner per quarter. We integrate the Raku Capture scan flow into your onboarding (white-label or sidecar web app), run roughly 100 customer scans, and co-publish the results. No charge for the pilot. We ask for UX-team access and permission to publish a case study.

What the pilot covers: scan-flow integration, MCP itemization against your preferred inventory format, a carrier review portal (so scan data stays on RakuAI cloud), and completion + adjuster-touchpoint metrics. Early pilots target InsurTechs first — SOC 2 Type II is in-flight for larger-carrier procurement.

Pilots run on RakuAI cloud with a carrier review portal until SOC 2 lands. No SOC 2 today — InsurTechs first.

Go deeper before you scan or write.

The pages most people read next.

Raku Capture →

The consumer scan flow this is built on — scan a space, then talk to it.

MCP Server →

The MCP surface that lets an LLM walk a scene and itemize what's in it. Production since March 2026.

Spatial Engine →

The C++ runtime under the splat reconstruction — the same core, no new engine for insurance.

Privacy →

Scanning a home is sensitive. How capture data is handled, and how to delete a scan.

Microsoft for Startups

RakuAI is an NVIDIA Inception member. The same NVIDIA-accelerated 3D Gaussian splatting stack we ship in Raku Capture is what reconstructs the room and powers the contents itemization described above. Read more →

NVIDIA Inception is a free program for AI startups; RakuAI's membership does not imply that NVIDIA endorses RakuAI's products. Learn more at nvidia.com/startups.

RakuAI is built on AWS and an AWS Activate participant. Production reconstruction runs on Azure today; AWS GPU capacity (G5 / G6e) is on our roadmap, honestly gated until quota lands.

“AWS” and “AWS Activate” are trademarks of Amazon.com, Inc. or its affiliates. RakuAI’s participation does not imply that AWS endorses RakuAI’s products.