Contracts and agreements
You sendLeases, supplier agreements, engagement letters.
You getThe dates, notice periods, renewal terms and liability caps, as a table your team can check.
01 · What it does
What you can do with it
The same jobs teams send to US-hosted models, done on a machine in Burnaby. Send text, get structured answers back, keep the files where they are.
Client files and case notes
You sendIntake forms, correspondence, meeting notes.
You getA summary and the open questions, without the file ever leaving the country.
Tickets, emails and forms
You sendWhatever lands in the inbox or the web form.
You getEach one tagged by topic and urgency, routed to the right person, drafted reply included.
Records that disagree
You sendSupplier lists, product data, spreadsheets from three systems.
You getMatched, de-duplicated and normalized records, in the JSON shape your database wants.
02 · Why Canada
Why it matters that it's in Canada
Neither OpenAI nor Anthropic processes requests in Canada. The usual advice for Canadian organisations is to go through a US hyperscaler's regional deployment, or buy a server and run a model yourselves. This is the third option: a model already running here, behind an API you can use this afternoon.
It helps when any of these are on your desk:
- PIPEDA makes you accountable for personal information you hand to a third party, wherever it goes.
- Quebec's Law 25 requires a privacy impact assessment before personal information leaves the province.
- Health information under PHIPA and its provincial cousins has its own rules on leaving the province.
- The US CLOUD Act lets US authorities compel US companies to hand over data they hold, wherever it is stored.
- Law and accounting firmsPrivileged and confidential files that can't go to a US cloud.
- Clinics and health servicesNotes and forms with personal health information.
- Financial and insuranceClient records under OSFI or provincial rules.
- Public sector and educationProcurement that asks where the data goes, and needs a plain answer.
Your application
Sends a prompt with the OpenAI SDK, base URL changed.
encrypted (TLS)
Cloudflare
Routes the encrypted request. Keeps no copy.
private tunnel
AltaMind, Burnaby BC
The model answers in memory. Nothing is written to disk; only token counts are kept, for billing.
Whether this satisfies a given regulation is a question for your privacy officer; the Data Processing Addendum has the details they'll ask for, and I'm happy to answer their technical questions.
03 · Integration
Two lines of code to switch
from openai import OpenAI client = OpenAI( # 1. the URL base_url="https://api.altamind.ca/v1", # 2. your key api_key="sk-alt-…", ) r = client.chat.completions.create( model="altamind-chat", messages=[{"role": "user", "content": "…"}], )
Works with what you already use
- The official OpenAI SDKs for Python, Node, .NET, Go and Java, and anything else that lets you set a base URL: LangChain, LlamaIndex, Vercel AI SDK, n8n, Make.
- Streaming, system prompts, multi-turn chat and JSON answers.
- One model name,
altamind-chat, so model upgrades never break your code. - Keys per client, revocable in seconds.
GET /v1/usageshows this month's count.
04 · Pricing
One fixed price a month
A generous allowance, no per-token meter, no surprise invoice. In Canadian dollars, plus applicable taxes.
Starter
$99CAD / month
About 30,000 pages of text a month
- 25 million tokens a month, in and out combined
- 2 requests at once, 60 a minute
- Internal tools, document runs, a first production job
- Key in your hands right after checkout
Business
$299CAD / month
About 130,000 pages of text a month
- 100 million tokens a month, in and out combined
- 4 requests at once, 240 a minute
- Production workloads and large document backlogs
- Upgrade from Starter any time; the difference is prorated
Cancel any time from your billing portal; you keep access to the end of the month you've paid for. Not a fit in the first 14 days? Email me and I'll refund the month in full. Want a pilot before paying? Email info@altamind.ca for a pilot code and enter it at checkout.
Subscribing means you agree to the Terms of Service. The Data Processing Addendum sets out what happens to personal information you send, for your privacy officer.
Need the whole machine, or a specific model?
A dedicated box in BC, running an open model of your choice, usually as part of a consulting engagement.
Questions people ask before subscribing
What's a token, and how many pages is 25 million?
A token is about three-quarters of a word. A full page of text is roughly 750 tokens, so 25 million tokens is about 30,000 pages a month, counting both what you send and what comes back. A typical contract of 12 pages, summarized, costs about 10,000 tokens.
What happens when I reach the monthly allowance?
Requests get a clear "limit reached" answer until the first of the next month. Nothing extra is ever charged. If you need more, switch to Business in the billing portal and the new allowance applies immediately.
Do you store my prompts or the answers?
No. They pass through memory and are never written to disk. What is kept per request: the timestamp, token counts and how long it took, so you can be billed and so I can see when the machine is struggling. The Data Processing Addendum spells this out, including where the data goes and who else touches it.
Which model is it, and can I choose?
Currently Gemma 4, an open-weight mixture-of-experts model that is fast at extraction, classification and summaries. It is served as altamind-chat, and when a better open model comes along I switch to it without changing your code. If you need a specific model, that's the dedicated option.
Is there an SLA?
No, and I'd rather say so than pretend. It's one monitored machine that restarts itself after a power cut. It suits internal tools, document runs and batch jobs. If your customer-facing product can't tolerate a short pause, talk to me about a dedicated setup or use it as a second, in-Canada option beside another provider.
How do I change my plan, update my card or cancel?
From Stripe's billing portal, linked on your welcome page and in the confirmation email. Upgrades apply right away with the difference prorated; downgrades and cancellations take effect at the end of the paid month. Receipts arrive from Stripe by email.
Who is behind this?
Me, Zhouwu Huang, the engineer who built and runs it. AltaMind Technologies is registered in British Columbia (FM1119348). If something's wrong you email me and get me, not a ticket queue.
05 · Contact
Not sure it fits?
Send a few lines about what you'd send it and roughly how much, and I'll tell you plainly whether this is the right tool, or whether something else is.
info@altamind.ca
Burnaby, British Columbia
Canada
Canada