Answered by the operator who runs it.

Two to four weeks for a single location, depending on how clean the data in your current system is. The cutover itself happens overnight — last close on the old system Sunday evening, first open on CannAgent Monday morning. The ramp before that is data audit, hardware swap, and three days of on-floor support during the first week. Multi-location chains stage cutovers one store at a time; we don’t flip ten registers at once. We have run this off Dutchie on our own floors, and the shape is the same coming from POSaBIT, Treez, Flowhub, Cova, BLAZE, Greenbits or BioTrack — what changes is how much of your history the outgoing vendor lets you export, which is the first thing we check.

No. We pull transactions, customers, loyalty points, vendor records, and inventory snapshots out of your current system before the cutover and load them into your new Postgres database. You see your last twelve months of revenue, your loyalty members keep their points, and your vendor history carries over. Anything we can’t migrate cleanly we tell you about up front — we don’t silently drop rows.

METRC continuity is scoped per state during onboarding. The manifest and transfer mapping is built and tested; the live feed connects per store as each operator’s METRC credentials clear — so METRC states (CO, MI, OK, MA, MO, NV, OR, NJ, IL and more) are in build today, not live. Washington runs its own system instead, CCRS, and we build that file natively. BioTrack states — CT, NM, DE — are regulatory coverage right now, not a connected feed. Some states are first-class, others ship as a custom integration on the Enterprise tier, and we tell you which bucket your state lands in before you sign anything.

No. Most operators we talk to are mid-contract with Dutchie, Treez, or Flowhub — that’s normal. We can run a parallel-environment migration where you stay live on the current POS through your renewal date, and we stage your CannAgent environment in the background. The actual cutover is a Sunday-night flip; your contract end-date doesn’t need to align with onboarding finish. If you want to hold off the cutover until renewal to avoid double-paying for two months, that’s the most common path. If you’d rather eat the overlap to switch sooner, that’s also fine — we don’t bill until you’re live.

No. Pricing is per location, billed monthly. Add a register, add a budtender, add a manager — same line on the bill. Cannabis cash-margin is tight enough without a vendor metering you on every staff add. Card-processing fees, if you turn them on, pass through at processor cost; we don’t take a per-swipe markup.

Yes — one base at $499 per location per month, flat, with unlimited registers and $0 setup. Modules like the online menu, loyalty, and payroll are flat add-ons you pick. At 2+ locations the base drops to $399 and add-ons come down 20%. The number we say in the demo is the number on the invoice, and your rate locks to your scope for life.

Month-to-month after month 12. The initial agreement is twelve months — that’s how long it takes us to absorb the cutover cost without charging a migration fee — and after that you’re month-to-month with 30 days’ notice to cancel. We don’t auto-renew you into another twelve. We don’t hold your data hostage on the way out: you get a full SQL export of your Postgres database, a CSV of your transactions and customer list, and your loyalty balances in the standard import shape. If we’re not earning the bill, you should leave.

Hybrid. Every page in the back office has an in-app help panel — 150-plus surfaces — so a new manager solves their question without texting the owner at 11pm. The training portal ships with role-tracks for budtenders, leads, inventory, purchasing, and admin. On top of that, every cutover includes a 30-minute operator-team onboarding call and three days of on-floor support during the switch. We don’t bill for training the people who run the store.

Most budtenders are comfortable on the register inside one or two shifts — the cart flow looks like the cart flow on every other cannabis POS, with the vertical-ID gate and the till-variance ladder as the noticeable adds. Inventory leads take a week or two to absorb the cycle-count, waste-log, and self-learning ordering surfaces. Owners get the back office (payroll, write-ups, performance reviews) as a sequence — we don’t hand you eleven new modules on day one.

No — the platform is built to cannabis retail rules, but it is not a HIPAA-covered system. Medical-patient verification is built in (patient ID checks, registry lookups, age-and-eligibility gating at the cart) and that flow is handled under state cannabis rules — in Washington, WAC 314-55-080 (medical cannabis endorsement) and chapter 246-71 WAC (the medical cannabis authorization database, including its confidentiality rule at 246-71-100) — not 45 CFR 164. PHI in patient ID-checks doesn’t cross into our system as covered HIPAA data. If you operate a medical-only retail surface that requires a BAA, that’s a separate conversation — talk to us at the demo.

Yes — WAC 314-55-155 governs cannabis advertising in Washington, and the menu / marketing surfaces inside CannAgent are scoped to it. The one operators most often miss is subsection (7): except for outdoor advertising, ALL advertising must carry four warning statements — intoxicating effects and habit forming; impairs concentration, coordination and judgment, do not drive or operate machinery; there may be health risks; and for adults 21 and older, keep out of reach of children — each at no less than 10% of the largest type in the advertisement. A text message is advertising and is not outdoor advertising, so all four apply, and they run 341 characters before your own copy begins. Outbound SMS templates, customer-engagement email and case-card content sit behind that block plus the (2)(a) content prohibitions and the (2)(c) 21+ statement. The product helps the business; we don’t make claims about cannabis itself, and the templates we ship don’t either.

You do. Always. Each location runs on its own Postgres database, exportable any time in standard SQL. No vendor-database lock-in. On the Enterprise tier, source-code escrow is available so the platform itself keeps running on your terms if anything ever happens to us. Your customer list is not aggregated, not resold, and not used to train any model — it sits in your database.

Each location runs on its own Postgres database hosted on Neon, with the application served from Vercel’s edge. The register has an offline cache for the last-known cart state, so an in-flight transaction completes even if the connection blips. A full regional outage on either provider would degrade the back office (analytics, reorder queue) before the register; we’ve held that as the design boundary from the start, because the register is the one surface a sale can’t wait on.

Claude (Anthropic) is the primary model behind the assistive features — write-up drafting, scope generation, the reorder-queue explainer. The AI proposes, the manager gates the action — see /admin/manager-writeups for AI-suggested proportionate-response options against WA labor + WSLCB rules, then a human acknowledges and ships. We don’t auto-fire anyone, auto-purchase from a vendor, or auto-run payroll. The model is assistive, not autonomous.

Only the minimum slice needed to draft the suggestion, and never your customer list. The reorder-queue explainer sees product SKUs and sell-through rates — not who bought them. The write-up drafter sees the incident facts and the relevant WA labor rule — not employee personal info. The scope generator sees what the prospect typed into the demo form — not your historical transaction set. Anthropic’s paid API tier doesn’t retain prompts for training — that’s their published commitment, not a handshake — and we don’t mirror prompts to any analytics tool. Patient PHI on the medical surface never crosses into AI calls. If your compliance team needs the per-feature data-scope written out, ask at the demo and we’ll send it.

Yes. Each location runs on its own Postgres database under its own state license, with the same workflow across all of them — so a parent LLC manages multiple stores without a separate tenant per vendor. Cross-store reporting and cost-aware tiers are part of the Multi tier; cross-state federation is Enterprise.

We’re developing the patent strategy with experienced cannabis IP counsel. Until filings are on record, we describe the implementation in the demo — not on a marketing page competitors can scrape. The black bar is the implementation detail held for the call. If you book a demo, you see everything — the self-learning vendor reliability + auto-PO scaling, the training-completion gate at the POS, the variance escalation ladder with the offset-window camera-review trigger, and the jurisdiction-aware proportionate-response engine on manager write-ups. We’d rather show you on a working call than print it for the next vendor to copy.

You keep your data exports and walk — and since setup and onboarding are free, there’s no fee sunk on the way in. We’d rather not work with someone unhappy. The full cross-tier guarantees — compliance built in, no migration fees, named human support, free training, your data stays yours — live at /pricing.