Parcel briefs that survive a counsel review.
Parcel briefs with a complete steward provenance chain attached to every value — built for Ohio's 88 counties.
For brokers who need to defend a number, land-bank directors who need to triage a backlog, and probate attorneys who need a single page their client can read.
Steward sources
- FEMA NFHL
- USGS 3DEP
- USDA SSURGO
- ODNR
- County GIS
Screening reference only — not a flood determination, wetland jurisdictional study, insurance quote, or suitability opinion for any specific use.
Provenance chain
parcel #159535 · Hocking, OH · sample
Every numeric field in every brief resolves to one of these three rows — the steward who owns the data, the ingest run that produced it, and the sha256 of the source artefact. No override path, no manual fill-in.
Coverage
88 counties.
Four touched.
One pattern that scales.
Strata88 is named for Ohio's 88 counties. Today it runs an active pilot in Hocking (founding rural pilot) with partial, lab-grade implementation in Franklin (urban-scale validation), Vinton (Tier 2 land-bank case), and Athens (lab onboarding). Every new county slots into the same catalog-attributed pipeline — no per-county code changes.
Why these counties?
Hocking is the founding rural pilot where the catalog originated, and where the ODNR-adjacent steward sources (oil & gas wells, soil-class, NFHL floodway) matter most. Franklin is the urban-density stress test — most steward sources, highest parcel volume, hardest naming-collision cases. Vinton is the Tier 2 land-bank case — the county where the diligence-diff and land-bank-triage workflows were designed. Athens is the newest lab county, partially implemented as the next expansion candidate.
- Franklin
- Columbus
- Urban county pilot
- Hocking
- Logan
- Founding pilot
- Vinton
- Mcarthur
- Tier 2 land-bank pilot
- Athens
- Athens
- Lab county (partial)
Boundaries · Ohio DOT (ODOT) county boundaries · all 88 counties shown
What you actually get
A token-gated brief, rendered server-side.
Every value is steward-attributed. Every claim is counsel-tiered. The link is revocable and the audit log records every fetch.
Parcel brief · Pilot
Parcel
159535
Hocking County, Ohio
Last refreshed
Wed 2026-05-27 · 14:23 EDT
Regulatory Floodway intersect (FEMA NFHL)
Regulatorily-unbuildable subset of SFHA. Confirm via FIRM and floodplain administrator.
Steward: fema_nfhl
Screening summary
- hit
Modeled flood zone (SFHA)
FEMA NFHL overlap heuristic — not LOMA or insurance rating.
- hit
Stream / flowline intersect
Large-scale NHD flowline geometry intersects parcel.
- clear
Wetland polygon overlap (NWI)
NWI mapping — field verification may differ.
- clear
Public / protected land overlap (PAD-US)
PAD-US overlap — not fee title or easement legal status alone.
What we checked
| steward | run | refreshed | artefact sha256 |
|---|---|---|---|
| FEMA NFHL | #100 | 2026-05-27 | 2b91…c4af |
| USGS NHD | #47 | 2026-05-26 | 6d04…91ce |
| ODNR oil/gas wells | #75 | 2026-05-27 | f1b2…8a73 |
| ODOT TIMS | #32 | 2026-05-24 | a78d…4e15 |
algorithm: diligence_diff_notifier_v1 · Screening reference only — not a flood determination, wetland jurisdictional study, insurance quote, or suitability opinion for any specific use.
Illustrative render of a Hocking County demo parcel. Every value resolves to a steward, an ingest run, and a source-row sha256.
Architecture
Three architectural choices that show up in every brief.
Without them you have one more parcel viewer. With them you have something a buyer, a regulator, or a counsel review can actually defend.
Counsel-clean by design
Every field in every brief is gated by a commercial-maturity tier and a counsel-hold flag. Claims that have not been cleared simply do not render — there is no override path. The platform refuses to lie on your behalf.
Steward-attributed every claim
Every overlay value resolves to a named steward (FEMA NFHL, ODNR, USGS, county GIS), the specific ingest run that produced it, and the source artefact sha256. The chain ships in every brief — no hand-waving, no "trust us" boxes.
Auditable from MCP tool to source row
Strata88 published its own MCP audit conformance standard (open source, sibling repo). Every agent-facing tool emits a v1 envelope so downstream callers can verify provenance without trusting Strata88's own server.
On every Land Card
Brokers interrogate the brief.
The bot only answers from the tool chain.
Six fixed playbooks call the same MCP tools that produced the brief itself — no free-form chat, no hallucinated stewards, no answers the audit log can't reproduce.
Pick a playbook
- Explain screening flags
- What wasn't checked?
- Explain soils
- Explain terrain
- Where did this data come from?
- Full screening brief
Each playbook runs a fixed MCP tool order. Same JSON the Land Card uses — no parallel truth, no second story.
Broker asks
"Where did each of these screening flags come from? I need to be able to cite this to my buyer."
Land Bot · provenance_receipts
3 tools · 412msThe five screening fields on this parcel resolve to the following stewards and ingest runs. Every value is reproducible from the run id and the source-row sha256.
| column | steward | run | refreshed |
|---|---|---|---|
| flood_zone_flag | fema_nfhl | #100 | 2026-05-27 |
| stream_intersect | usgs_nhd | #47 | 2026-05-26 |
| wetland_overlap | usgs_nwi | #21 | 2026-05-24 |
| soil_hydric_share | ssurgo_oh | #88 | 2026-05-25 |
| elevation_aspect_mode | usgs_3dep | #15 | 2026-05-20 |
Tool chain: parcel_provenance_chain_tool
· returned 5 rows · screening reference only — not a
flood determination, wetland jurisdictional study,
insurance quote, or suitability opinion for any specific use.
What it refuses to answer
- "Is this parcel a good investment?" — no commercial outcome promises.
- "You're flood-free." — no binary determinations; flags + escalation only.
- "Can I build a cabin here?" — lists relevant flags, defers to zoning / surveyor / Corps.
Illustrative render of the in-product Land Bot panel. Playbook names match shipped LB0–LB3 playbooks in the BULU repo; sample values are representative, not live. Reference: LAND_BOT_PANEL_IMPLEMENTATION_PLAN.md.
Who it's for
Built for three concrete jobs.
Strata88 is opinionated about who it's for. Pick the role closest to yours — every other claim on this page follows from there.
A counsel-clean parcel brief you can hand a buyer in two minutes.
One token-gated link contains the screening summary, the steward-attributed overlay table, and a printable record. Your client opens it on their phone — no signup, no account, no marketing follow-up.
If anyone questions the brief later, every value resolves to a named steward, the ingest run that produced it, and the source row sha256.
Request a sample brief
Triage the Tier 2 backlog without paying a consultant.
A free batch script ranks your held inventory against the same screening overlays Strata88 uses internally — flood, wetland, slope, soil-suitability. Runs on your own machines; no upload, no PII required.
Get the Tier 2 triage script
Probate attorneysHand heirs a single page that explains what they inherited.
Heir's Liability Shield briefs summarise the screening flags and steward provenance for an inherited Ohio parcel. You keep the attorney–client relationship; Strata88 is the screening reference your closing letter cites.
Request a sample heir's brief
Engineering
The discipline IS the product.
Strata88 is a one-person operation today. The architecture choices, test coverage, and stewardship metadata are the substitute for having a team to enforce them — and the reason a counsel review passes the first time instead of the fifth.
- 1,097
- test functions, green on every PR
- 345
- Python modules in the active app
- 46.5k
- lines of Python
- 54
- Alembic migrations
- 37
- MCP tools (+ 7 prompts, 11 resources)
- 0
- TODO / FIXME markers in active code
Stack
- FastAPI
- PostgreSQL · PostGIS · pgvector
- Celery · Redis
- Alembic
- Jinja2
- MCP (Model Context Protocol)
- Astro · Tailwind · React
- Cloudflare Pages
Source
Disciplines that show up in every brief
- Strict type discipline
- ruff + mypy strict across every source file. The full test suite is gated in CI on every PR; nothing merges red. Refactors are mechanical because the types and tests catch them.
- Catalog stewardship metadata
- Every data source carries a YAML manifest with steward class, data plane, commercial gate, and rotation cadence. CI fails if the manifest drifts from the catalog. No silent dataset additions.
- Lifecycle-tiered documentation
- ADR-governed (ADR 0007 + 8 ADRs total) docs grouped by lifecycle (active, runbook, archived) and tier (core, ops, archive). The TOC for 137 active docs is auto-generated from front matter and CI-checked for drift.
- Open MCP audit conformance standard
- Strata88 published its own MCP audit envelope (v1) as a sibling repo so downstream agents can verify provenance independently — without having to trust Strata88's own tool implementations.