Every lease. Every unit. One graph.
TenantGraph turns tenancy paperwork into verified, source-linked records, then runs the working day from one cockpit: money, obligations, flow and assets. Built only from your own documents. No scraped listings, no third-party data.
Numbers you can check, mechanically
AI reads your documents, but nothing it says is taken on faith. Three checks run on every extracted value, and anything that fails waits for a person. A green mark comes from evidence, never from model confidence.
verified Verbatim quotes
Every extracted value carries a source quote that is matched against the document’s own text. Open the evidence before you confirm the record.
“£1,725 per calendar month” · p.1needs review Values must live inside their quote
A genuine quote attached to the wrong number still fails. Amounts and dates must appear inside the cited text, or the field drops to human review.
Illustrative mismatch: £1,850 ✕ “£1,725 per calendar month”computed Arithmetic outranks opinion
Rules with a computable reading, such as England’s deposit cap, are calculated from human-confirmed fields. Not a model’s view: a number you can re-derive.
Sable Row: £3,000 deposit · £2,769.23 calculated cap · £230.77 overA few seconds inside the working preview
Follow a letting team through its day: find a clause, file incoming evidence, review an application and prepare a maintenance response.
Find the wording before you reply.
King Street requires a written pet-consent request. The answer cites the clause on page 2, so you can open it before replying to the tenant. The wording does not say consent has been given.
Give every document a place.
Daniel’s payslip and bank statement arrive together with a Parkside lease and a utility bill. Clear matches enter their workflows. The utility bill waits in the inbox for a person to file.
Spot the mismatch. Review the evidence.
Daniel’s application contains a bank statement that belongs to Maya. The filing mismatch and the rent-to-income check stay visible until a person resolves them, before any decision is made.
Read the report. Choose the next step.
Riverside reports no hot water. Read the description, the attachment and the access windows, then apply the suggested priority. The diagnosis and the visit still need a person to confirm them.
Three layers. Nine modules. One data spine.
Adopt a single module or the whole platform. Every module reads and writes the same records, so nothing is entered twice.
Document Intelligence
Every document becomes structured, verified data.
- Lease Extraction45-field profile, every value quoted
- Compliance Review35 rules in two packs, clause-quoted findings
- Document VerificationFive material types, cross-checked without AI
Operations
Daily portfolio work, run from one cockpit.
- Rent Ledger & ArrearsMonthly rent roll, live balances per lease
- Deposits & Compliance CalendarReceipt to return; gas, EICR and EPC expiries
- Work OrdersPhotos, visits and AI triage a person applies
Communication Intelligence
Every conversation answered from your records.
- Portfolio Q&A in preview buildCited, verified answers from lease documents
- Tenant NoticesDrafted from the record, sent by a person
- Owner ReportingPortfolio summaries built from the ledger
What the preview build does today
Beyond the recordings: the working features already in the build, grouped by where they sit in your day. Everything below runs end to end against the demo portfolio.
Run the day
-
Operations cockpit
One screen on sign-in: money collected against money due, obligations, application flow and assets, plus a needs-attention queue with one-click actions.
-
Workbench triage
Everything waiting for a decision in one list, with an inspector beside it. Drive it from the keyboard: j k to move, Enter to act, next item pre-selected.
-
Case file per lease
One dossier per tenancy: applicant, application, lease and tenants on a relationship ribbon, plus one timeline across rent, deposit, compliance and tickets.
-
Search, actions and answers
⌘K from any page finds leases, applications, documents and tickets, offers the actions that fit the moment, and answers a question with citations inline.
-
Batch inbox
Drop up to 20 mixed files. Leases go to extraction, applicant materials attach to the matching application, and anything ambiguous waits for you. The classifier routes; it never fills in data.
Keep the record
-
Applicant to tenant
Approve, reject or withdraw an application with a note and timestamp, then link the approval to the signed lease. The decision travels with the tenancy.
-
Lease lifecycle
Renew, end early or reopen. Status derives from the term (upcoming, active, ended), events stay inside it, and every change is written to an audit table.
-
Deposits, step by step
Record receipt, protection with scheme and reference, and return with itemised deductions. The protection deadline lands in the event queue.
-
Rent ledger and arrears
A monthly rent roll built from committed leases. Settle or void each due event and see the live outstanding balance per lease, with its payment history.
-
Compliance calendar
Gas safety, EICR and EPC certificates per property, validity periods applied for you. Expiries join the same queue as rent and renewals.
-
Work orders with AI triage
Tickets carry photos and scheduled visits. The model suggests priority, trade, urgency flags and questions for the tenant. A person applies them, or not.
Make it yours
-
Your fields, checks and rules
Switch the 45 stock lease fields on or off, add custom fields with extraction hints, set the application checklist and rent-to-income ceiling, and choose the generic or England rule pack.
-
Bring your own model key
Anthropic or OpenAI models behind one interface, with your own API key encrypted at rest and switchable per account. A content-aware mock provider runs demos at no cost.
-
Model bench
Run the same document through a second model and compare values and evidence checks side by side. The primary record is never touched.
-
Exports and activity trail
Leases and the rent ledger as CSV, every original document opened from the record it produced, and an activity timeline of what happened across the account.
Modules are views. The graph is the asset.
Five entities, property, unit, lease, tenant and event, form one connected spine. Every module reads and writes the same records, so nothing drifts out of sync.
TenantGraph is not a shared SaaS. We deploy it into a Cloudflare account you own, with your portfolio in its own database, then hand over the keys and roll out updates. Nothing is shared with other customers and nothing leaves your account.
It sits in your own account beside whatever you run today. Start with one module, import what exists and export at any time: leases and ledger as CSV, documents as originals. No rip-and-replace.
Decisions, payments, deposit steps, certificates, renewals and ticket changes appear on one activity timeline, with who did what.
Access keys with sessions, file types checked by content rather than filename, duplicate uploads caught by hash, and a job queue with retries and dead-lettering so nothing sticks in “processing”.
Anthropic or OpenAI behind one interface, encrypted at rest and switchable per account. No key means no silent fallback: an unconfigured provider fails visibly.
Everything that must happen becomes an event in one queue: rent due, renewal, deposit deadline, certificate expiry, visit. That queue is the product’s heartbeat.
Now onboarding a first cohort of design partners: rental operators managing 200–5,000 units. Each partner gets a deployment in their own Cloudflare account, set up by us. Shape the roadmap and keep founding terms.
See your portfolio as a graph.
A 30-minute walkthrough with your own documents. We assemble a live, source-traceable record from one lease, in front of you.