BuildMindHQ is a multi-tenant CRM for construction businesses. Pipeline, estimates, price book, pre-construction handoff, the job workspace and invoicing all hang off one job record — and the whole thing ships under your company's name, not ours.
Desktop-first, because that is where estimates get written. Six primary sections, twenty-six tool-belt shortcuts, one AI assistant that follows you across all of them.
A job does not start as a job. It starts as a phone call and ends as a cleared payment. Each stage below is a real part of the app, not a diagram.
A drag-and-drop pipeline board with nine stages, from New Lead through Signed to Completed or Lost. Inbox, tasks and calendar sit alongside it, and the dashboard counts what is where.
Assemblies built from your own price book, priced cost × (1 + markup) with the resulting margin shown on every line. Flip to client view to hide costs, export a quotation PDF, or turn the estimate straight into an invoice.
Generate a handoff package from the project and send it as a link. The recipient opens it without an account, sees only what you chose to publish, and acknowledges it — which is recorded against the job.
A per-job workspace with timeline, materials, checklist, files and a crew Q&A thread. Field notes go in as text and come back out as a structured daily log with milestones, materials and flagged concerns.
Invoice numbers are issued by the server, so two people cannot mint the same one. Payments are appended atomically. A Stripe Checkout link is generated per invoice when your Stripe key is set.
Everything on this page is described from the code, not from a roadmap. Where something is partial, it says so — see build status.
Pipeline counts, pipeline value, active jobs and tasks due are queried straight from your projects and tasks tables. Where the app has no real data model behind a panel — build schedule, mentions — it renders empty rather than inventing a number.
Owner, admin and staff. Invite, promote, demote and remove — all gated server-side, with the last owner protected from removal. Every one of those actions writes to an activity log automatically.
A searchable catalogue by category with cost, markup, price, unit and vendor of record on every item. Margin is computed, not typed, and flagged when it drops below your floor.
Today, overdue, future and undated — as a board or a list, with counts on each. The month grid places the same tasks on their due dates, so nothing has to be kept in two places.
Numbers are issued by a database function rather than by whoever clicked first. Payments append atomically. Where a Stripe key is configured, each invoice can produce a real Checkout session.
Every core table carries an organisation id and row-level security written against it. One company cannot read another's rows even if the application code forgets to check, because the check is not in the application code.
The pre-construction package goes out as a link. Whoever receives it opens it in a browser, sees only what the package contains, and acknowledges it — which comes back and lands on the job.
Scoped to the shell rather than a page, so the conversation survives navigation. It reads your job context, never invents a price or a code citation, and marks anything uncertain for a human to check.
BuildMindHQ was built multi-tenant from the schema up, not retrofitted. Each company that signs up gets its own organisation, its own logins and its own colour.
One house rule runs through all of it: never invent a price, a part number or a code citation, and mark anything uncertain so a human checks it. Each module needs an API key configured, and fails closed rather than guessing when it is missing.
Point it at a site photo and it reads the scope against your price book — never a generic catalogue — and returns a low/high range with a confidence badge and its reasoning. Every request is kept so you can see what it saw.
Limit Input is an image URL today. There is no camera or file picker yet.
Give it an address and a project type and it researches permit requirements, setbacks, HOA notes and code references with live web search, then writes a checklist.
Limit Output is a draft to verify before filing. It does not file anything with any jurisdiction.
Paste what the crew reported and get back a structured entry: summary, milestones, materials used, flagged concerns, and a separate plain-language draft for the client.
Limit Text in, text out. Audio recording and transcription are not implemented.
Each AI feature costs a fixed number of tokens from your organisation's balance, deducted atomically before the call goes out. If the request then fails on our side, the tokens go straight back. Nobody pays for a request that never returned an answer.
It falls back to a fixed local reference set — bonding, drain standards, pressure tests, cure and startup, barrier code — rather than producing a plausible answer with nothing behind it. That is a smaller answer, and an honest one.
Six sections you live in, and a tool belt of twenty-six shortcuts underneath. On a phone both rows collapse into one menu; the app is designed for the desk it gets used at.
Not all of these are equally finished. The next section says which is which, screen by screen, and we would rather you read that before a demo than after one.
BuildMindHQ is in active development. Some of it is wired to a real database and some of it is still a screen with sample data behind it. Here is the split, so nothing in a demo surprises you.
| Area | State | What that means |
|---|---|---|
| Accounts, organisations, roles | Live | Signup, onboarding, password reset, invites, promote/demote/remove, activity log. Isolation enforced by row-level security. |
| Pipeline, tasks, calendar, price book, estimates | Live | Reading and writing real rows once a backend is connected. |
| Invoices and payment links | Live | Server-issued numbers, atomic payment recording, real Stripe Checkout sessions when a Stripe key is set. |
| Pre-construction handoff | Live | Generate, send, and a public token link the recipient opens without an account. |
| Dashboard | Partial | Pipeline, active jobs and tasks due are real. The funnel, lead-source and revenue charts and the weekly digest are still sample figures — treat them as layout, not data. |
| AI modules | Partial | Built and metered end to end, but each one requires an API key to be configured and fails closed until it is. |
| Job workspace | Partial | Timeline, materials, checklist, files and crew Q&A read real job data. The Specs, P&L, Gallery, Subs, Permits and Time tabs are sample content. |
| Email delivery | Not built | Nothing is sent by email yet. A team invite writes a record; "send to client" on an invoice changes its status. Neither puts a message in anyone's inbox. |
| Client portal, client read-tracking, sub-account manager | Not built | These screens exist and are populated with sample content. There is no live portal, no open- or download-tracking, and the sub-account list is illustrative. |
| Site plans, architectural plans, renderings | Not built | Placeholder screens only. Blue Tide runs these in separate applications today; whether they ever land inside BuildMindHQ is an open product decision. |
There are no customer counts, testimonials or results figures anywhere on this site, because there is nothing honest to put in them yet. When there is, it will be here with a date on it.
It shares infrastructure, and it is separated by the database rather than by application code. Every core table carries an organisation id, and the row-security policy on each one filters by the calling user's organisation. A query for another company's rows returns nothing. The privilege paths that could move a user between organisations are handled by specific database functions with their own permission checks — direct column edits are revoked outright.
Your company name and primary colour are stored on your organisation record and applied across the interface when the app loads. Logo upload and custom domains are in the schema but not wired into the interface yet — ask about timing rather than assuming them.
It says so. The rule enforced in every prompt is that it never invents a price, a part number or a code citation, and marks anything it is unsure of for a human to verify. The Vision Estimator prices only against your own price book. The Permit Drafter labels its output a draft. If the assistant cannot reach the model at all it falls back to a fixed reference set instead of improvising.
It is readable and usable on a phone — both navigation rows collapse into one menu below 600 pixels. But it is designed desktop-first on purpose. Estimating, price-book work and pipeline management are desk jobs, and pretending otherwise produces software that is bad at both.
Pricing is not published yet, and self-serve subscription billing is not built — plan tiers exist in the schema but nothing charges a card for them today. Onboarding is hands-on for now. Email us and we will talk about what your setup would need.
Yes. Export and full organisation deletion are handled as requests today rather than as a button in the interface — write to privacy@buildmindhq.com and see the data deletion instructions for exactly what happens and how long it takes.
A walkthrough runs about thirty minutes on a shared screen. Bring one of your own jobs and we will put it through the pipeline, price it out of a price book, and show you where the software stops.