Identity and location contracts
Define the read-only identity, workspace, resource, and exact-location contracts used by the MVP.
Exact head: not yet reported
Founder working visual · MVP proof 18% · July 14, 2026
The neutral acceleration layer between property managers, their data, and every tool or AI agent they choose.
We make fragmented operating data trustworthy, portable, safe to use, and ready to power better decisions. Clients keep running property management. We help accelerate everything around it.
Live MVP finish line
Finish the read-only identity, location, connection, and action contracts before the commercial MVP feature waves begin.
Required proof gates verified — not an estimate of time or effort.
Define the read-only identity, workspace, resource, and exact-location contracts used by the MVP.
Exact head: not yet reported
Freeze read-only connection semantics while keeping source mutation outside the MVP.
Exact head: not yet reported
No delivery blocker is reported.
PR #88 and PR #89 are isolated, dependency-ordered, and limited to synthetic read-only contracts.
PR #85 passed its protected merge and post-merge gates.
Founder-approved direction
All 22 MVP proof gates, final evidence reconciliation, and the founder go or no-go come first. This roadmap does not change the MVP denominator or its verified progress.
Founder approval sets direction. Each conditional product advances only after its visible evidence gate is satisfied. No dates or percentages are implied.
The company thesis
There will not be one permanent software winner in property management. The stack changes too quickly. The durable position is the independent data and acceleration layer that lets clients adopt the best tools without rebuilding their business every time.
PMS, pricing, operations, guest messaging, finance, CRM, documents, spreadsheets, media, email, Slack, and future tools.
Identity, history, evidence, decisions, permissions, APIs, agents, exports, and beautiful information surfaces.
Owner intelligence, onboarding, content, websites, social, CRM, compliance administration, bookkeeping support, and capabilities not invented yet.
Business model
The original premium price was based too heavily on delivery cost and consulting analogues. The corrected model counts only value that can reach client cash flow. A typical 75 property manager is usually too small. Healthy operators near 250 to 300 properties are the practical center unless a smaller company has a major migration, hiring, acquisition, or risk trigger. Lake Tahoe Accommodations is a design partner candidate, not proof of the final market.
Software COGS · July 14
COGS here means software, infrastructure, APIs, and third-party vendor charges only. Taylor allocates founder time, human labor, onboarding, support, and overhead separately.
$10.35/property$207 software floor
Four-property proving shape. Human delivery is excluded; unresolved source and shared-runtime charges remain outside the known subtotal.$0.49/property$292.91 software floor
Workers, R2, and Supabase Pro/Micro only. Five unresolved contract or invoice items are shown below instead of being guessed.$0.54/property$1,337.79 software floor
Shared, read-only, standard-helper case. This is a total portfolio floor, not a per-client quote.What Paid edge runtime, APIs, scheduled dispatch, queues and native logs.
Why Runs the control plane and ingestion dispatch. LTA's modeled 302,400 monthly queue operations remain inside included allowances.
Notes Conservative full platform allocation; incremental LTA cost may be $0 on the existing paid account.
What 240 GB of raw evidence plus immutable payload/document writes.
Why Makes canonical data replayable and preserves provenance independently of Postgres.
Notes $3.45 storage + $9 Class A operations after free allowances and billing-unit rounding. Target architecture; not yet deployed.
What Pro organization, isolated Postgres project, daily backups, Auth allowance and one $10 compute credit.
Why Stores canonical operational data, contracts, tenant authorization and indexes.
Notes Micro is approximately $10 and is covered by the credit. No $100 PITR and no Small upgrade are needed for the replayable read-only pilot.
What Current Hostinger runtime used by the existing system.
Why Some persistent/background jobs may still run there until displaced or measured.
Notes Incremental LTA cost is currently $0. The shared allocation cannot be verified without the actual invoice and utilization.
What Streamline, Breezeway, Airbnb/Vrbo partner rights, RevA identity and any direct PriceLabs access.
Why Access rights—not compute—are the likeliest first material cost or bottleneck.
Notes Keep client subscriptions client-paid. PriceLabs direct API would be $120/month at this size; Conduit would be $1,499/month and is excluded unless deliberately bundled.
These are activated by explicit reliability, workload or enterprise requirements—not included by default.
The displayed floors use an 80% software gross-margin target. They are not final commercial prices.
Taylor independently sets founder time, onboarding, support, account management, risk, and final pricing.
The first wedge
Before dashboards or AI accelerators can be trusted, the company needs one reliable property knowledge layer. The first product proves that scattered facts can be captured, reconciled, certified, surfaced, and exported without losing history.
Map every client's unique sources and terminology into a wide, extensible data model.
Keep business information current across PMS, pricing, operations, AI, documents, and human managed sources.
Combine Streamline, PriceLabs, RevA, Conduit AI, Breezeway, Slack, and other systems into a real time owner view.
Reuse governed data for listings, captions, social content, websites, CRM, bookkeeping support, and later workflows.
Approved architecture
The commercial wedge stays narrow. The underlying data foundation must be wide enough to support new clients, sources, verticals, AI models, and services without redesigning the platform.
Clients, resources, identities, source objects, evidence, claims, decisions, relationships, artifacts, policies, approvals, receipts, and audit events.
Clients and verticals add typed resources, fields, relationships, units, validation, sensitivity, authority, and namespaces without redesigning the database.
Permanent client owned IDs. Vendor IDs are mappings, never the identity of the business.
Complete exports. Facts, evidence, history, relationships, artifacts, permissions, and receipts leave with the client.
Open surfaces. Stable APIs, events, webhooks, agent tools, and connector contracts support whatever the client builds next.
Controlled repointing. New destinations receive explicit commands, diffs, approvals, readbacks, and rollback receipts.
The client surface
We provide a beautiful standard scorecard and a controlled customization layer. Clients can rearrange it themselves, ask Claude to propose a unique version, or build through our API without forking the platform or bypassing data policy.
Supported identity, data health, freshness, conflict, provenance, access, and audit components.
Layout, labels, branding, filters, thresholds, audience views, and approved metrics remain portable and versioned.
Every change receives a preview, policy check, cost estimate, compatibility test, approval, and rollback point.
AI safety and private compute
The platform should encourage clients to build with their own or our AI agents, while making safe behavior the default. For the most sensitive workloads, local models can operate inside the client's controlled environment so frontier labs never receive the data.
Approved public facts and published content.
Operational context with purpose and user checks.
Minimized, redacted, or routed to client controlled compute.
Local models and purpose bound access only.
Every execution is attributed to workspace, actor, workflow, purpose, provider, model, tokens, tools, storage, local compute, and payer. Platform-funded, client-funded, and bring-your-own agents share budgets, warnings, hard stops, per-run ceilings, model restrictions, and auditable escalation rules.
Competition and moat
They own core workflows and can expand into analytics or integrations. We win by remaining portable across them and making migration less dangerous.
They provide infrastructure but usually do not deliver the STR specific canonical model, evidence, stewardship, operating controls, and acceleration partnership.
They consume or create operational information. We make that information governed, safe, portable, and reusable across agents.
They can solve workflows manually. We must turn repeated work into a reusable client controlled foundation instead of accumulating labor.
Execution sequence
Inventory OwnerRez, Breezeway, Google Drive, Docs, Sheets, Supabase, Notion, Conduit, Airbnb, Vrbo, the direct site, onboarding files, guidebooks, manuals, email, Slack, local files, and every forgotten source. Unknown and unconnected sources remain visible.
Create permanent identities, immutable capture, the evidence vault, policies, decision history, and complete export without production writeback.
Reuse the proven fact engine, dossiers, claim tracing, approvals, write guards, readback verification, and receipts. Wrap vendor specific code as adapters.
Run the new canonical model beside current systems, compare every fact and artifact, preserve disagreements, and time the human work.
Show sources, freshness, evidence, conflicts, unknowns, inferences, certifications, published diffs, and a complete export in a beautiful owner ready surface.
Use certified facts for listings, captions, Conduit knowledge, and the operating dashboard. Add controlled writes only after explicit approval, canary scope, readback, receipts, and rollback.
Certify baselines and reconstruct history before any new system writeback. Repoint one consumer at a time only after parity and correctness are proven.
Roam Free acceptance gates
Every known source inventoried. Unconnected sources remain visible.
Every active property has a permanent ID and verified external mappings.
Selected source captures are immutable, hashed, and replayable.
At least 95% of the target fact catalog is populated or explicitly unknown or conflicted.
Every displayed or exported value traces to evidence and a decision.
Authority, sensitivity, audience, freshness, and retention are defined for every target field.
Restricted data egress to external models fails closed.
No unresolved critical safety, access, policy, or identity conflicts.
A complete export and independent restore both succeed.
All setup, review, correction, and maintenance time is measured.
Evidence room
Capability manifests, client agents, token spend, scorecards, portability, security, organizational separation, and the frozen MVP.
Read all 44 decisionsSoftware-only portfolio costs, vendor bottlenecks, gross-margin floors, buy-versus-build choices, and the explicit owner-allocation boundary.
Read corrected briefVendor thresholds, source constraints, deployment modes, failure costs, scenario ranges, and independent formula review—with human overhead excluded.
Audit the modelThe governing technical blueprint, data lifecycle, safety model, migration path, and MVP gates.
Explore blueprintWhat to reuse, generalize, wrap, retire, and build without rewriting proven code.
Read auditProduct fit, competition, SWOT, target clients, moat, risks, and validation roadmap.
Read reportBuyer value, delivery economics, Standard scope, qualification, and kill criteria.
Read reportVeeva, Athenahealth, Celonis, ServiceTitan, failure patterns, and first ten client gates.
Read reportMarket alternatives, implementation accelerators, standards, safety, and client readiness.
Read reportWhat worked in healthcare, restaurants, trades, construction, employment, finance, and ecommerce.
Read reportGo directly from the analysis to original filings, documentation, standards, company material, and reporting.
Browse sourcesHow the finish line moves
Open work is visible immediately, but the percentage changes only when a frozen proof gate is fully satisfied. That keeps the board useful during fast parallel delivery without turning activity into fake completion.