Privacy
What this site records, and what it does not
Operated by Tempo AI LLC, accountable for the calculation, publication and correction of every figure published here.
There is no account, no wall and no advertising here, so there is very little to record. What follows is not a policy written next to the code: it is the code. The page count, server-side traffic observations and advisory need submissions described below have their source in the public repository.
The page count, field by field
Every page carries one line of script that calls /api/t once. That request records exactly six things, and nothing else:
| Field | Value | Why |
|---|---|---|
| type | one of site_visit, cli_query, mcp_call, share | tells a page view apart from a machine read |
| target | the page or category, at most 80 characters | which study is read at all |
| ts | the time, to the second | whether readership grows or stops |
| surface | site, cli or mcp | whether readers are people or agents |
| source_class | agent, human or unknown, from the user-agent string | same question, coarser and never stored raw |
| country | the two-letter country the edge reports | where the readership is, never who |
Server-side traffic observations
Middleware also logs the UTC time, normalized route (without query strings or route identifiers), claimed request class and provider, crawl or fetch purpose, evidence kind, origin-verification state, recognized assistant referrer hostname, collection source, internal-test flag and coverage limit. Only recognized assistant hostnames are retained; referrer paths, other referrer hostnames and raw user-agent strings are discarded. Middleware cannot see the eventual response status, so that field is unknown. The hosting provider separately records request status in its own logs.
Browser-like traffic is not proof of a person. Crawler names and assistant referrers are unverified claims, not proof that an assistant cited TORNEO. Requests bearing our x-torneo-internal test header are excluded from audience readbacks. Middleware logs have partial coverage: cached/static traffic, missing logs and untagged internal tests prevent complete visitor counts. These observations are kept separate from the browser page count, with no visitor identifier or cross-page linking. Page-count beacons, static assets and Next.js prefetch/RSC requests are excluded; aggregate rows keep the normalized route so API requests remain distinguishable from page requests.
The current hosting plan is Vercel Pro, with documented baseline runtime-log retention of one day (15 October 2026). Authorized aggregate readbacks may be retained as dated operational evidence. No analytics provider, cookie, raw IP address or new personal identifier is added by this traffic instrument.
Advisory tournament requests
When you use submit_need or request_tournament through the API, CLI or MCP, TORNEO logs the submitted task category, requester kind and opaque origin, optional constraints, priorities, adaptation and description, with a content-derived need identifier. Do not submit personal information or secrets. These requests do not start runs. Runtime log retention is limited by the hosting provider; a logged request is not a guarantee of durable storage. The response states this limitation.
Retained request events may be imported into a private queue. Its public snapshot exposes only the need identifier, advisory state and any linked run, with no submitted text or requester identifier. Follow-up states appear after ingestion and publication. Quote your need identifier for a request concerning that stored submission; the contact below applies.
What is not recorded
The page count uses no cookie, no local storage and keeps no IP address. No visitor identifier, not even a hashed one: an earlier version derived a salted daily visitor hash, and it was removed in September 2026 because no written legal basis justified it. One consequence is stated plainly rather than hidden: returning readers cannot be told apart from new ones, and no figure published here claims otherwise. No name, no email address, no profile, no cross-page linking, no third-party analytics, no advertising network.
Who else sees a request
Vercel Inc. serves this site and holds its runtime logs for a short retention window; a request line there carries the path, the response code and the moment, and it is where the counts above are read from before being aggregated. GitHub, Inc. hosts the source repository, so following a source link takes the reader to GitHub under GitHub's own terms. Both are named as processors in the data protection packet that ships with the repository.
Measurement runs are a separate matter from this page: what each run collects, from whom, why, and for how long is declared per run in that same packet, and every published run carries its own legal preflight.
Your rights, and how to use them
Swiss data protection law applies, and the GDPR in addition for data subjects in the European Union. The page count described here is operated by Tempo AI LLC. Write to emea@devlo.ch, the contact point declared in the data protection packet, to ask what is held about you, to have it corrected or erased, or to object. Because this page count holds no identifier, we usually cannot connect a request to a person: that is a limit of the exercise, and the honest answer is that there is nothing to return.
Legal notice · Who operates this · What results cannot tell you