Pricing¶

The short version: we charge for compute we can't run on our own hardware. Everything we can run on the home server in Edmonton is free — that's most of the site.

The rule¶

Canadian Political Data is operated on a single, modest server hosted in amiskwaciy-waskahikan (Edmonton), Treaty 6 territory, behind a Canadian network connection. Most of the product runs there at no marginal cost to us:

The full Hansard text corpus and the embeddings index.
Politician rosters, ridings, bills, votes, hosting-sovereignty data.
Semantic search — including the AI model that converts your query into a vector. That model runs on a local GPU. It costs us electricity, not API fees.
The map, the explore view, alerts, saved searches, magic-link login.

When you use the public site, none of that compute leaves the building. It is also why we can keep the public side free indefinitely.

The pieces we can't run locally are the ones that need a much larger language model than our hardware can host. Today that's exactly one feature: generating a report. Reports are produced by a large hosted language model (via a third-party inference API) summarising the relevant speeches with citations. That call costs us real money, per report, in proportion to how much text the model has to read and write.

So reports cost credits. Search doesn't.

What's free, forever¶

Searching Hansard — full semantic search across every speech we've ingested.
Browsing politicians, ridings, bills, votes — the entire structured dataset.
The map and the explore view.
Saved searches and alerts — we run the comparison job ourselves on a schedule; no external compute.
Submitting corrections — and you earn credits for accepted ones.

We will not put any of these behind a paywall. If something currently free becomes paid in the future, it will be because the cost shape genuinely changed (e.g. we couldn't keep running it on local hardware), and we will say so explicitly.

What costs credits¶

Today: report generation. A report is an AI-assisted, evidence-cited summary of what a politician has said about a topic across Hansard, with every quote linking back to its source. The model that writes the summary is hosted by a third-party inference provider — there is a direct, per-token cost to us each time you run one.

Pricing details — pack sizes, per-report cost, the hold/commit/release flow, refunds — are in the Reports and credits section. The short version relevant to this page:

The cost is shown to you before you commit to running a report.
Credits are placed on hold, not spent, until the report succeeds. If it fails, the hold is released and your balance is unchanged.
If a report consumes less than the up-front estimate, the unused portion is automatically refunded to your balance.
Spent credits are not refundable in general; bug-driven failures on our side are. See Buying credits.

What might cost credits later¶

There are a few features on the roadmap that, like reports, are fundamentally bottlenecked on compute we can't host locally:

Programmatic API access for institutional users — bulk export, scheduled topic alerts, and "compare A vs B" tooling for journalists, academics, and advocacy organizations. The public web UI stays free; the paid tier covers the operating cost of serving high-volume programmatic workloads.
Voice / accessibility features that depend on hosted speech models.

When something new becomes paid, the same rule applies: the price exists because the underlying compute has a real per-call cost we can't absorb into the home server. We will say so on the page where the charge appears.

What will not cost credits¶

We are not going to:

Charge for the things the home server can serve at scale (search, browsing, the map, alerts).
Charge for "premium" politician profiles, "early access" data, or anything that's just a reshuffling of the public record behind a gate.
Run ads, sell user data, or take political-party donations or sponsorships. The public side is funded by credit purchases for the things we genuinely can't run for free, plus, where applicable, grants.

How the money flows¶

Credit purchases are processed by Stripe. We see a Stripe customer ID, the pack purchased, and the amount paid — not your card number. We pay the third-party LLM provider directly out of credit revenue. There is no external investor extracting margin from this product.

If you are a journalist, academic, educator, or community organization and the credit cost is a barrier to legitimate work, please get in touch — we grant credits for public-interest use.

Why it's structured this way¶

The alternative pricing models all looked worse:

Subscriptions — would require us to charge users a recurring fee to subsidise a feature most of them never use, and would push us toward locking up the free side to make subs feel valuable.
Ads or tracking — would compromise the "access without surveillance" stance and would not pay the bills at our scale anyway.
Charging for search — would defeat the point. The product exists to make the parliamentary record actually accessible.

Tying price to compute we can't run locally is the cleanest way to keep the public side free, the paid side honest, and the incentives aligned: we make money only when you ask the system to do something genuinely expensive on your behalf, and only after you've seen the price.