Sancera
Get started

Technology

How the technology works

Sancera sits between your AI assistant and the institutions' data services. This page describes what happens when you ask a question, what the assistant gets back, and what we store.

From question to answer

The flow from your question through the assistant and Sancera to the source and back.
  1. 01

    You ask a question

    You write as usual in Claude, ChatGPT or Cursor.

  2. 02

    The assistant calls Sancera

    The assistant decides the question needs data and calls one of Sancera's tools over MCP, for example with a series and a period. Sancera receives the call, not your conversation.

  3. 03

    Sancera retrieves from the source

    The data is retrieved from the institution's official data service, for example SCB or Riksbanken.

  4. 04

    The answer comes with its source

    The assistant receives structured data with source, retrieval time and terms, and writes its answer from that.

Sancera does not generate figures or alter source data.

Three steps in the large databases

SCB has more than 5,000 tables. In large statistics databases the assistant therefore works in three steps, and only the last one counts as a data request.

  1. 01

    Search

    The assistant finds the right table using plain words, in Swedish or English.

  2. 02

    Read metadata

    It reads which variables, regions and periods the table contains.

  3. 03

    Retrieve

    It retrieves the selection the question is about.

For common questions, such as the policy rate, exchange rates or GDP, there are ready-made tools that go straight to the right series.

What the assistant gets back

This is the response when the assistant asks for Riksbanken's policy rate. The example is shortened, but the fields are the real ones.

Response from Sancera (shortened)
{
  "series": "policy_rate",
  "description": "Riksbank policy rate (styrränta)",
  "unit": "percent",
  "latest": { "date": "2026-10-09", "value": 1.75 },
  "source": "Riksbanken SWEA",
  "source_url": "https://www.riksbank.se/en-gb/statistics/",
  "primary_source": true,
  "retrieved_at": "2026-10-11T15:03:18+00:00",
  "citation": "Riksbanken SWEA, data 2026-10-09, retrieved via Sancera on 2026-10-11. https://www.riksbank.se/en-gb/statistics/",
  "staleness": { "stale": false, "last_period": "2026-10-09", "cadence": "daily" },
  "license": {
    "name": "Riksbank API terms (FAQ) — free, commercial reuse allowed, attribution required",
    "status": "verified",
    "checked": "2026-09-20"
  },
  "export": {
    "excel_url": "https://mcp.sancera.ai/export/….xlsx",
    "csv_url": "https://mcp.sancera.ai/export/….csv"
  }
}
source, source_url
Which institution and data service the figure comes from.
retrieved_at
When Sancera retrieved it.
citation
A ready-made source reference the assistant can put in the answer.
staleness
The latest period in the series and whether it lags the source's publication schedule.
license
The source's terms and the date we last checked them.
export
Links to the same data as .xlsx and .csv.

Connection and keys

One address
The assistant connects to https://mcp.sancera.ai/mcp over HTTPS. Any MCP client that supports remote servers can connect.
Personal key
Every user has their own key. In a team each person has their own, and you can replace your key in the portal.
Sign-in without a password
You sign in to the portal with a one-time code sent to your email address.
Payment with Stripe
Card details are entered with Stripe and are not handled by Sancera.

What we store

We store what is needed for quota, billing and operations.

What we store per call
Which tool and which source was called, the time, the response time, whether the call succeeded, and which client it came from, for example Claude or ChatGPT.
What we don't store
We don't store the question itself. To improve source selection we keep, for 90 days, a de-identified template of the question with names, places, numbers and dates removed. The template is not linked to your account.
Your assistant
Your AI assistant is provided by a third party and processes your questions under its own privacy policy.
Read the privacy policy

Sources and terms

Sancera only uses open data from public institutions and their official data services, and only serves data whose terms allow it. The terms are included in every response.

See all data sources

Questions about technology or security?

Email hello@sancera.ai, also if you need a data processing agreement.

Get started free How to connect Sancera to your AI assistant