hoste · listing onboarding automation

Hoste already runs onboarding through Typeform, Guesty, Breezeway, PriceLabs, Besty, and team photography. Today the administrative path from form submit to a ready listing takes about one to two weeks. That is not because someone works that whole time. Data, photos, and decisions arrive in parallel and must be coordinated.

This proposal covers two paths: a Claude-assisted setup on a computer, or an owned pipeline (Render + Google Cloud Run Jobs + DeepSeek Flash) with a dashboard on Hoste's domain. The recommended path is the owned pipeline: lower recurring cost per listing, durable state across the multi-day cycle, and human effort concentrated on photography, pricing exceptions, and final approval.

A basic Claude Pro seat would not sustain the session volume of this operation. A realistic Claude path starts at Max 5x (US$ 100/mo) and can require Max 20x (US$ 200/mo). Even then, the main cost is administrative hours. The ROI case is hours freed per listing, plus a pipeline Hoste owns and can extend.

path a · claude on a computer. Configure instructions, content templates, and tools so an agent helps with listing setup. A person watches execution, unblocks issues, and approves publication.

• -setup + assisted pilot: US$ 280 (20h × US$ 14/h)
• -Claude Max 5x: US$ 100/mo baseline (Pro at US$ 20 does not hold the sessions)
• -Claude Max 20x: US$ 200/mo for heavier weeks · both monthly only
• -best for few listings, variable process, idea validation
• -no owned dashboard or autonomous integrations

Listing setup means long browser sessions and many tool calls per property. Pro runs out within a session. Max raises the ceiling but keeps five-hour and weekly caps, so no plan guarantees a fixed number of listings. Every run still needs a person at the computer, and the subscription does not cover the API a custom system uses.

path b · owned pipeline (recommended). Take the technical pipeline end-to-end with durable state, parallel wait points, and a dashboard for completeness.

All options side by side at 20h/week. Each bar ends at a working delivery, not at a document.

B2 runs the same phases. Weeks 2 to 5 stretch to cover browser automation wherever the Guesty API has gaps.

Once live, every new listing follows its own one to two week cycle. The system only spends compute when something arrives.

Onboarding is not one linear job. After the initial Guesty listing, Breezeway onboarding, photos, visit data, and commercial rules can advance in parallel. The system needs a coordinator that keeps one listing state open for one to two weeks and resumes when each delivery arrives. It does not hold a browser open while waiting.

completeness requirements for point-to-point coverage:

• -completeness · checklist of data, photos, configs, and integrations needed to publish
• -resume · continue from the last confirmed step after failures or restarts
• -idempotency · Typeform retries must not create duplicate listings
• -coordination · owners, due dates, blockers, alerts for stalled items
• -verification · confirm what was written in each system and that go-live happened

Guesty documents listing creation via Open API; the MCP server is currently read-only. We will confirm write permissions on Hoste's account in the first hours. Where API coverage is incomplete, Python + browser automation fills the gap.

Human effort stays where it creates value. The system presents what is missing, what diverges, and what needs a decision. It does not ask for every field again.

Photos are Hoste's team work. The pipeline waits for them and resumes. DeepSeek Flash handles labeling: interpret answers, organize amenities, produce titles/descriptions from confirmed data. Fee rules, dates, and publication stay in code.

No new domain purchase. Use something like dashboard.iamhoste.com (subdomain on the existing domain). A path like iamhoste.com/dashboard would require wiring into the current site. The subdomain is cleaner.

Simulation assumptions: ~20 minutes total compute per listing (including resumes), 2 vCPU / 2 GiB, region us-central1, and ~10k input + 2k output tokens for DeepSeek. Flash peak pricing without cache ≈ US$ 0.0054 per listing. Off-peak roughly halves the AI line. Numbers are planning estimates until the pilot measures real usage.

owned pipeline vs Claude subscriptions. same volumes, recurring cost only. Pro is left out because it does not hold the sessions.

The API alternative is not cheaper either. As an illustration only: a Sonnet-style agent accumulating ~500k input + 30k output tokens could spend ~US$ 1.95 in AI per listing (about US$ 195 at 100 listings), versus ~US$ 0.54 DeepSeek in the owned path. That agent figure must be measured before any decision.

The week-long cycle is mostly waiting. The expensive part is active administrative work: receiving information, filling systems, chasing blockers, and checking results. Automation should shrink that active work; photography and travel stay with the team.

illustrative comparison at US$ 14/h (same rate as this proposal):

This does not automatically remove a full salary. If the person stays, you free capacity for more listings or other work. Cash savings appear when you avoid a new hire, overtime, outsourcing, or other real spend. At the illustrative US$ 490/mo capacity release, a US$ 1,120 build pays back in roughly 2.3 months before infra and maintenance.

subscriptions vs build, at 100 listings/mo, before counting any admin hours:

The API-first build is already cheaper than Max 5x in the first year. Against subscription cost alone, it pays back in ~9 months vs Max 5x and ~4 months vs Max 20x. Browser-heavy pays back in ~15 and ~7 months. Over three years the API-first build saves ~US$ 2,560 vs Max 5x and ~US$ 6,160 vs Max 20x. That is before the admin hours saved, which are the larger number.

A Claude subscription rents capacity that runs on one person's computer. An owned pipeline is an asset: Hoste keeps the code, the data, and the integrations, and can extend them.

Once Typeform, Guesty, Breezeway, PriceLabs, and Besty are wired into one system with shared state, each new automation is a smaller project. The expensive part, connecting the tools and modeling the data, is already paid for. Possible next modules, each scoped and priced separately:

• -listing quality monitor · flag listings with missing photos, weak copy, or drift between channels
• -homeowner reporting · onboarding status and performance summaries sent to owners automatically
• -pricing & occupancy alerts · PriceLabs and calendar signals turned into actions for the team
• -operations follow-up · Breezeway tasks, cleanings, and maintenance tickets tracked against each property
• -reviews & guest messages · classify feedback and draft replies for approval
• -new markets · the same flow serves new Colorado and Arizona cities without extra headcount
• -other apps · Slack, email, Google Drive, CRM, or BI connected through the same webhooks and API

Claude can still be one of the models inside the pipeline where it is the best fit. The difference is that Hoste controls when it runs, what it costs, and what happens to the results.

in scope (owned pipeline)

• -Typeform webhook → validate → initial Guesty create/update
• -field fill, fees/calendar/messages where automatable
• -DeepSeek labeling + title/description generation
• -photo upload after team delivery · PriceLabs / Besty hooks as agreed
• -dashboard with login · history · blockers · resume · alerts
• -deploy, tests on real listings, publication after approval

stays with Hoste

• -site visits, photography, on-the-ground verification
• -pricing decisions, commercial exceptions, final go-live approval
• -existing SaaS plans: Typeform, Guesty, Breezeway, PriceLabs, Besty, storage, domain renewal

The SOP still has incomplete steps. Those must be defined before full automation. We automate available technical actions and coordinate human dependencies. We do not pretend every SOP line is already machine-ready.

• -maintenance · US$ 30/h when needed; optional 2h/mo reserve = US$ 50/mo
• -possible extras · plan upgrades, Guesty API access fees, external lookup services, taxes/FX if applicable
• -total company cost · current plans + automation + actual maintenance + human work + applicable charges

Start with the owned pipeline targeting US$ 1,120 · ~4 weeks at 20h/week, confirming the technical path in the first four hours. Operating cost stays between ~US$ 3 and ~US$ 60/mo across the simulated volumes, below a single Claude Max 5x seat. On top of that come the admin hours saved and a pipeline Hoste can extend. Both get measured in the pilot.

acceptance. One real listing completes the agreed flow: create, configure, reviewed content, verified integrations, publish after approval. That includes resume after an interruption without duplicating the listing.

To close ROI precisely we need three Hoste inputs: listings per month, admin hours per listing today, and monthly cost of the person who owns the process.