The permit record is public.
The intelligence isn’t.
Site Box measures permitting reality — what’s required, how long it takes, what fails — across all 39 municipalities of Palm Beach County SOURCED · live count, and indexes every permitting authority in the United States REGISTRY · 20,024 authorities. Developers, contractors, and lenders buy the answer before ground breaks.
Every number on this page tells you where it came from.
That’s the product. Competitors assert accuracy; Site Box shows provenance — on the homepage, in the app, and in every report a customer can put in front of a bank.
SOURCED = live database counts as of July 2026. REGISTRY = national index scaffolding, ahead of measurement.
One platform, three working surfaces.
All three run today on live infrastructure. Screens below are drawn from the working product; project figures shown are SYNTHETIC · demo — no real parcel is depicted.
| Cost band | $412k – $486k fees & exactions | T2 |
| P50 / P80 | 217 d / 288 d to vertical | T2 |
| Fatal flaws | 0 of 10 screens tripped | T1 |
| Gate | Human review · 2 decisions | HITL |
Feasibility verdict
A guided intake feeds an agent that resolves requirements against measured code, then a human-gated verdict: go, conditional, or no-go — with the cost band and clock attached.
LIVE — auth, billing & investor replay deployed| Site plan approval | DRC review | 3 blocking |
| ERP — SFWMD | submitted | on track |
| Utility capacity | evidence | verified |
| Exaction exposure | computed | $1.77M |
| Reg change feed | watching 7 agencies | live |
Approval command center
Once a project commits, Site Box becomes its system of record: the full approval pathway, an evidence locker with verification states, exaction math, and a regulatory change feed.
LIVE — bound to production data| Authority | City of Boca Raton — Development Services |
| Requirements | indexed · checklists published |
| Review clock | benchmarks pending grounding gate |
| Neighbors | /fl/palm-beach/delray-beach · /highland-beach |
Jurisdiction pages
A page for every state, county, and place in the country — full depth where we’ve measured, honest registry stubs where we haven’t. The public index is the acquisition engine and the demand sensor.
ARCHITECTED — national URL grammar, data live in registryAnswers with citations, not vibes.
Pre-rendered from measured Palm Beach County data — an illustrative demonstration of the in-product analyst. Every answer terminates in the code section or record that produced it.
A residential pool enclosure in Boca Raton typically requires a building permit with engineered screen-enclosure plans, plus electrical if lighting or bonding is touched. Within a flood zone, elevation documentation is pulled into the checklist automatically — Boca Raton has mapped zones on parcel record. Site Box assembles the exact checklist from the adopted code, not a generic template.
SOURCES · City of Boca Raton Code of Ordinances (indexed corpus) · FBC §454 / R4501 · geo.flood_zones parcel overlay
Both cities run commercial build-outs through plan review with fire and health touchpoints, but the review structures differ: West Palm Beach routes through combined development services review, while Jupiter separates building from planning-zoning sign-off — which changes resubmission dynamics. Site Box renders each jurisdiction’s actual sequence, then benchmarks the clock as measured outcome data clears our grounding gate.
SOURCES · WPB + Jupiter indexed code corpora · measured review sequences · timeline benchmarks: pending grounding gate (we don’t publish what we can’t defend)
Re-roof inspections in the county cluster around a small set of failure modes: fastening-pattern documentation, underlayment verification, and missing product approvals (NOA/FL numbers). Site Box links each failure mode to the requirement that triggers it, so a contractor fixes the paperwork before the truck rolls.
SOURCES · golden.permit_inspections (measured corpus, PBC) · requirement linkage via ARL · rate stats publish after out-of-sample gate
Provenance is the product — and the barrier.
Anyone can wrap an LLM around a municipal website. Site Box built the discipline layer: three orthogonal scales, enforced in the schema, rendered in the UI, and human-gated before anything is asserted.
Coverage — how deep we are, per jurisdiction
- C1 MEASURED outcomes measured: timelines, inspections, requirements
- C2 HARVESTED code & requirements extracted; outcomes pending
- C3 REGISTRY authority identified & geocoded; page says so honestly
Confidence — how each datum is grounded
- T1 PRIMARY the adopted instrument itself, cited by section
- T2 VERIFIED verified secondary evidence, dated
- T3 INFERENCE model output — always labeled, human-gated
Requirement grades — feasibility coverage
- V VERIFIED requirement resolved against primary source
- R REPORTED resolved against reported practice
- A ASSUMED assumption — surfaced, never silent
Depth where it exists. Breadth as trajectory.
Expansion is a data-depth problem, not a rebuild: every jurisdiction page runs one template, and a county promotes itself C3 → C2 → C1 as harvest and measurement land. No new frontend per market — ever.
The C3 index is also the demand sensor: “request coverage” clicks tell us which market to measure next — expansion sequencing by revealed demand, not intuition.
Two boxes, one flywheel.
sitebox.build is the community — construction news, topics, and contractor conversation for the market we serve. sitebox.ai is the platform. The community feeds the platform customers; the platform feeds the community authority; and everything users say in either box becomes product intelligence.
The feedback layer is built into every product surface — rate an answer, flag a requirement, request a jurisdiction — and routed through the same intelligence pipeline that mines the community. Product development is sequenced by revealed demand from working contractors, not by roadmap theater.
Tiered usage over one credit ledger, managed on the spread.
Everything Site Box sells — a verdict, an analyst answer, a month of monitoring, a title-risk screen, whatever the demand loop surfaces next — is metered in credits with a known per-action cost-of-goods. Customers experience simple subscription tiers with usage multipliers and opt-in on-demand overage; the business underneath manages one spread across every product at once.
Site Box Community
A sitebox.build profile, topics and answers, limited analyst queries, and “request coverage” voting. The top of the funnel — and the demand sensor.
For the working contractor
Enough monthly credits for routine verdicts and analyst sessions, one live project workspace, full jurisdiction depth where measured.
For the busy GC / developer
Multiplied credit allotment, multiple project workspaces, full command center with monitoring and change-feed revalidation, priority human verification.
For firms & power users
Maximum limits, team seats, early access to new intelligence products (title risk, exactions) as they clear grounding gates, API preview.
Beyond the plan’s included quota, customers opt into list-price credits — default off, so a bill is never a surprise. Overage is the highest-margin revenue in the system and it scales with customer success.
Tier prices and multipliers shown are directional PLANNED · pre-launch pricing — final anchors set against measured cost-of-goods per credit. The invariant is the structure: allotment tiers over one metered credit ledger, margin managed on the spread, every future product priced in the same currency.
The numbers we won’t show you yet
Our determination engine scores F1 99.66% — in-sample IN-SAMPLE · out-of-sample gate in progress. Until the held-out validation gate clears, that number appears here with this label and nowhere in marketing. Review-time benchmarks ship the same way: measured, gated, then published.
A company selling ground truth cannot round up. The discipline that slows our marketing is the discipline a customer’s lender will eventually audit — and that is precisely the bet.