Skip to content
ThingWeaveKeep projects moving on evidence

Legal

Privacy Policy

This page explains what ThingWeave ("we") collects on thingweave.com, the signed-in workspace at /account and the agent endpoint at /mcp, what it is used for, which third parties process it, and what you can see or change yourself. Last updated: 23 October 2026.

Information from Google sign-in

You sign in with Google. We request only the openid, email and profile scopes. From the signed ID token Google returns, we keep:

  • your Google account ID (the sub value), which is the identifier your projects are linked to;
  • your email address, only if Google reports it as verified;
  • your name as provided by Google (up to 200 characters), used as your display name.

Our sign-in code does not store Google access tokens, refresh tokens or ID tokens, and we do not request access to Gmail, Google Drive or any other Google service. Projects are linked to your Google account ID, not to your email address.

While you are signed in, a session record in our database holds a hash of your session token, your Google account ID, email, name, and the session's creation and expiry times. Sessions expire after 8 hours; signing out deletes the session record. During sign-in, a temporary record (valid for 10 minutes) holds a one-time nonce, a PKCE verifier and the page to return to.

Information you or your agent submit

When you use the workspace, directly or through an AI agent you authorize, we store:

  • project titles, the free-text requirement description (brief) you submit, and structured requirements: purpose, budget, quantity, target delivery date, specifications, constraints and materials;
  • each submitted version of the requirements and brief, and snapshots of the project state so earlier versions can be restored;
  • project context records such as goals, questions, answers, decisions and corrections, research tasks, research results, plans and your decisions on them;
  • requests you open on a project (questions, changes, cancellations, acceptance issues, aftercare) and their description;
  • attachments you approve one by one (currently UTF-8 plain-text .txt files only): the file name, size and SHA-256 fingerprint, and the file content;
  • agent connections: the client name the agent reports, the permissions you granted, expiry and revocation times, and a log of project actions (for example "requirements submitted");
  • rate-limit and usage-limit counters linked to your account ID.

Please do not put information in a project that you are not allowed to share with us or with the agents you connect.

AI agents you connect

When you authorize an AI agent (for example over MCP, the Model Context Protocol), we store the agent client's registration (its self-reported name and redirect address) and the authorization grant. The grant carries your Google account ID, email and name so the agent's requests act as you. Access tokens are valid for 5 minutes. A refresh token rotates on every use, ends after 30 days without use, and always ends within 90 days, when you need to approve the agent again. An authorization created before this limit does not gain that longer life: it ends seven days after its first refresh under the new limit. You can revoke an agent from the account page; we keep a record of the revocation.

An agent you connect can read the projects you authorized and write within the permissions you granted. What that agent and its provider do with the data it reads is governed by their own terms and privacy policy, not ours.

How we use this information

  • to sign you in and keep you signed in;
  • to store your projects and show them to you and to the agents you authorize;
  • to enforce access control, rate limits and capacity limits, and to keep the service secure;
  • with your separate consent only, to measure our public pages and our Google Ads, as described below.

We do not sell personal information. Our code does not send project content, attachments, your name or your email to Google Analytics or Google Ads, and they are not sent to manufacturing suppliers or other third-party businesses, except to the AI agents you choose to connect (which read the projects you authorize, as described above). They are stored and processed by the infrastructure providers listed below. Payments are not currently enabled, and the service has no payment form.

Google Analytics and Google Ads

Our public pages (home, product, use cases, workflow, cases, docs and blog) load the Google tag for Google Analytics 4 (G-J0V73DMVH9) and Google Ads (AW-18495510094). We use Google Consent Mode in its advanced form. Before the tag loads, all four consent types (ad_storage, ad_user_data, ad_personalization and analytics_storage) are set to denied. Until you allow something, the tag runs with all consent denied, which under Google's Consent Mode means it should not set cookies; Google still receives basic cookieless measurement signals (for example a page view), which include the network and device information needed to make the request.

A consent bar on those pages offers two independent choices, both off by default:

  • Analytics. If you allow it, Google Analytics may set analytics cookies (which we configure to expire after 30 days) and counts views of our public pages and clicks on fixed links.
  • Ad measurement. If you allow it, Google Ads may set advertising measurement cookies. On the account page, an isolated page with no project content then tells Google Ads when you create a project, submit structured requirements, or make your first submission. Each event carries a fixed value of 1 SGD and a de-duplication ID (a random project or submission ID, or a SHA-256 digest of your account ID). Our code only sends events from the last 7 days.

Our code always sends ad_personalization as denied, and we do not use this data for personalized ads or remarketing. The page address our code gives Google Analytics keeps only Google's ad click identifiers (gclid, gbraid, wbraid, gad_source) and validated campaign (UTM) parameters, so Google can tell which ad or campaign a visit came from; other query parameters are removed from it. For the page you came from, our code passes only the other site's origin (for example https://www.google.com), never its path or query, and nothing when you move between our own pages. The Google tag itself also reads the page address and referrer from your browser: requests that the tag sends to Google Ads on its own (such as page-view and consent-update requests) can include the full address of the page, with all of its query parameters (for example a fbclid added by another site), and the referrer your browser reports (browsers normally report only the other site's origin, and our pages tell browsers not to send our own addresses as a referrer). While cookies are not allowed, the Google tag may add ad click identifiers and a temporary Google Analytics identifier (the _gl parameter) to links between our pages instead of using a cookie, and until you allow ad measurement it redacts ad click identifiers from Google Ads requests. Our own systems do not store click identifiers; if you allow ad measurement, a Google advertising cookie on our site may hold one. Our code does not configure or send measurement if your browser sends Do Not Track or Global Privacy Control, if you are signed in to ThingWeave while viewing a public page, or on any host other than thingweave.com and www.thingweave.com. This Privacy Policy and our Terms page do not include the Google tag.

You can change or withdraw either choice at any time in the privacy choices panel at the end of our public pages (for example Docs) and, for ad measurement, on the account page. Data Google has already received is handled under Google's Privacy Policy (see also how Google uses information from sites that use its services).

Cloudflare hosting

ThingWeave runs on Cloudflare. Our application runs on Cloudflare Workers, and your data is stored in Cloudflare D1 (database), Cloudflare R2 (attachment files) and Cloudflare Workers KV (agent authorizations). Cloudflare processes request information such as IP address and browser details to deliver and protect the site. Pages served through Cloudflare also currently include Cloudflare's Web Analytics script, which measures page loads and performance. See Cloudflare's Privacy Policy.

Our code can also keep coarse daily counts of sign-ins, agent requests, requirement submissions and uploads, without names, emails, project IDs or content. This is currently switched off.

Infrastructure providers

These providers process data on our behalf to run the service:

  • Cloudflare: hosting (Cloudflare Workers) and storage: the database (D1), attachment files (R2) and agent authorizations (Workers KV);
  • Google: sign-in (Google account), and, as described above, Google Analytics 4 and Google Ads measurement on public pages and the account page.

Cookies and browser storage

ThingWeave itself sets these cookies, both Secure and HttpOnly:

  • __Host-thingweave_login: links your browser to a sign-in in progress; expires after 10 minutes;
  • __Host-thingweave_session: keeps you signed in; expires after 8 hours or when you sign out.

Our pages also use your browser's storage:

  • local storage thingweave.analytics-consent.v1 and thingweave.ads-consent.v1: your two consent choices and when you made them; after 180 days we ask again;
  • local storage thingweave.ads-sent.v1: only after you allow ad measurement, the de-duplication IDs of conversions already sent from this browser (the last 500);
  • session storage thingweave.probe.v1: a flag for our own testing, set only when a page is opened with our internal test parameter or when the browser is one of our own probes; cleared when the tab closes. Turning that parameter off clears the flag, unless the browser is still one of our probes.

Our configuration lets Google set its cookies only after you allow the matching choice, as described above.

How long we keep information

Sign-in records and sessions expire as described above. Project content, requirement versions, attachments and related records are kept until they are deleted; there is currently no automatic deletion schedule for them. Revoking an agent connection or a file permission stops further access but does not delete records already saved.

Seeing, correcting or deleting your data

When signed in, you can see your projects, requirements, attachments and agent connections on the account page, download research and plan assets as plain text, and revoke agent connections.

To ask for a copy of your data, a correction, or deletion of your account data:

Email privacy@thingweave.com. This address handles privacy questions and requests to access, correct or delete your data.

Changes to this policy

ThingWeave is a pilot and this policy will change as the service changes. When it does, we will update this page and its date.

See also our Terms of Service.