Rootline Privacy Policy
Last updated: November 15, 2026
1. About Rootline
Rootline (https://rootline.site) finds and explains failed runs of AI agents. "We" and "us" in this policy mean the operator of Rootline, identified in section 7.
We expect most of our users to be in the United States, United Kingdom, Canada and Australia.
Our role. For your account data, we are the controller. For trace data you send or connect, you are the controller and we process it only to provide Rootline to you (we are your processor / service provider).
2. Data we collect
Account data (you give us)
- Name and email address.
- Password. We store only a one-way hash; never the password itself.
- Email preferences (alert settings) and the AI explanations on/off setting.
- When you sign up, the date and time you accepted this policy and the version (its "Last updated" date), so we can show you agreed.
- Your workspace name and plan (Free, Team or Agency).
Integration credentials (you give us)
- If you connect Langfuse: the host URL, public key and secret key. The secret key is encrypted before storage.
- Langfuse API keys give access to your whole Langfuse project. Rootline uses them only to read traces and observations (our Langfuse client only makes read requests); we never create, change or delete data in your Langfuse project. You can revoke the key in Langfuse at any time, which immediately stops our access.
- If you send traces via OpenTelemetry: we issue you an ingest key and store only a hash of it.
Trace data (from your AI agents)
- When you connect Langfuse, we read your traces and observations (the last 30 days at first, then new ones about every minute). When you send OpenTelemetry data, we store the spans you send.
- Trace data can include whatever your agent processed: user prompts, model outputs, tool inputs and outputs, error messages, timings, token counts, costs, model names and tags. This may include personal data about your own end users. You control what your agent logs and sends to us.
- Recommended: mask personal data before it reaches us. Rootline only sees what your agent logs. If you use Langfuse, its client-side masking lets you remove or hash PII before traces are stored, so Rootline never receives it.
Data we generate
- Failure detections, rule-based and AI-written explanations, cost summaries, and records of alerts we sent (to avoid duplicates).
Technical and security data
- Session records (random token, expiry) and one-time verification / password-reset codes (stored hashed, short-lived).
- Your IP address is used in memory to rate-limit login and password-reset attempts; it is not written to our database. Our hosting provider may keep standard server logs, including IP addresses.
- On public pages and at a few product milestones (demo open and first interaction, signup start, signup form interaction, signup submission (including whether the attempt could not be completed, without recording the values you typed), signup complete, data-source connection started and completed, first trace received, checkout started, login), we record cookieless first-party events on our own servers. Each event is a page or action name, an optional UTM/source tag from the link you arrived on, and standard request metadata (IP address used only in memory for abuse control, not stored with the event; browser and referrer where the server receives them). These first-party events work without a cookie, and we do not set advertising cookies. Only if you accept analytics cookies do we also set a random first-party visit identifier and a 30-minute visit cookie (section 11), so a returning visitor is counted once and pages viewed in one visit are grouped together; if you decline or make no choice, neither is set. We also keep a record of your cookie choice on our own servers: when you click Accept or Decline on our cookie banner, we store which one you chose, when, and the page you were on. If you accept, the record includes the random visit identifier, so we can count how many visitors accepted and link the page you were viewing to that identifier. If you decline, the record contains no identifier, browser string or referrer, so it cannot be linked to you or to your other visits. To tell real visits from automated traffic, we also record whether a page showed a sign of a real visit (scrolling, a click or key press, or staying open and visible for about 10 seconds), and we mark, rather than delete, events from automated tools and from our own team so they do not skew the counts. One of these checks compares your IP address, in memory, with the address ranges that cloud-hosting providers publish; a match marks the event as likely automated, and the IP address is still not stored. People using a VPN or a company network hosted in such a cloud may be marked this way; it only affects our statistics, never your access. These events are aggregate product metrics, not a profile of you across sites. To understand which channels bring us customers, we also note where a visit came from (only the referring website’s domain, such as a search engine or AI assistant, and any UTM tag), the page it started on, and a pseudonymous visitor code computed from your IP address and browser with a random value we discard daily, so the code cannot be linked across days; if you sign up, we keep that first-visit source with your account, together with whether you later connect a data source or start a paid plan.
- Product analytics (only if you accept analytics cookies). If you click Accept on our cookie banner, our website loads a product analytics script, served from our own server, that sends usage data to a third-party product analytics service provider whose servers are in the United States. It sets a first-party analytics cookie on our domain and records the pages you view, how you arrived (including UTM/source tags), clicks on public pages (never what you type into fields), browser, device and approximate location derived from your IP address, so we can understand how visitors use the site and improve it. It uses the same random visit identifier as our own first-party analytics (section 11), so the two can be counted together, and it does not create a profile of you unless you are signed in. With the same consent, it also records session replays of how our pages are used (mouse movement, clicks, scrolling and the page layout as you see it), so we can find where the site is confusing or broken. Everything you type into form fields is masked in these replays. On signed-in pages its events are page views only (no click events, and page titles are replaced with a generic one), and in replays all text and element details are masked and trace content is left out, so your trace data is never recorded or sent to the analytics provider. Web addresses are stripped of query strings and record ids before they are sent. If you are signed in, we link these events to your internal Rootline user id (never your name or email) so we can see how the product is used across visits. Visits by our own team are not sent. If you click Decline, or make no choice, the script is not loaded and no analytics cookie is set. If you decline after accepting, the script stops recording and sending data, and its cookie and the data it stored in your browser are deleted. You can change your choice at any time with “Cookie settings” at the bottom of our public pages.
Payment data
- Paid plans are billed by a third-party payment provider acting as merchant of record. The provider collects your payment details, billing name, email and address and processes the payment; we do not see or store your full card details. We receive and store your subscription status, plan, a customer and subscription reference, and a link to manage your subscription.
3. How we use data and legal bases
- To provide Rootline (sync and store traces, detect and explain failed runs, show costs) and run your account (sign-in, verification, password reset). Legal basis (UK/EU): contract.
- Service and requested emails: verification and reset codes, and loop / cost alerts on Team and Agency plans (each has a one-click turn-off link). Legal basis: contract; legitimate interests for service notices.
- AI explanations of failed runs (section 5), which you can turn off. Legal basis: contract.
- Security, abuse prevention and support (section 6), and aggregate cookieless first-party product analytics. Legal basis: legitimate interests. Product analytics through our analytics service provider (analytics cookie and session replays, section 2), and the first-party visit identifier (
rl_vid/rl_sid, section 11), run only if you accept analytics cookies. Legal basis: consent, which you can withdraw at any time. Keeping a record of your cookie choice (accept or decline): legitimate interests, so that we can show that we asked for your choice and respected it. - If we ever send marketing emails, we will ask for consent where the law requires it.
We do not sell your personal data or share it for cross-context behavioural advertising. We do not use your trace data to train AI models.
4. Service providers (sub-processors)
We use a small number of trusted third-party service providers to run Rootline. They process data only on our behalf. By category:
| Provider type | Purpose | Data involved | Location |
|---|---|---|---|
| Cloud hosting and database | Hosting of the app and its database | All data in section 2 | USA |
| AI provider | AI-written failure explanations (unless you turn them off) | Redacted, truncated excerpt of a failed run (section 5) | USA |
| Email delivery | Sending emails (codes, alerts) | Your email address and email content (which may summarise failures and costs) | USA |
| First-party product analytics | Cookieless event counts on public pages and product milestones (demo open and first interaction, signup start, signup form interaction, signup submission (including whether the attempt could not be completed, without recording the values you typed), signup complete, data-source connection started and completed, first trace received, checkout started, paid plan, login), including optional UTM/source | Event name, UTM/source, page, referring domain, daily pseudonymous visitor code, whether the page view was engaged; your cookie choice (accept or decline, and when); a random visit identifier only if you accept analytics cookies; first-visit source kept with your account if you sign up; IP not stored | USA (our hosting) |
| Product analytics | Understanding how visitors use the website (page views, funnels, session replays), only if you accept analytics cookies | Pages viewed, clicks on public pages, session replays (mouse movement, clicks and scrolling; everything typed is masked, and all text is masked on signed-in pages), browser and device, approximate location, the random visit identifier, internal user id if signed in; no form inputs, no trace data | USA |
| Payment provider | Payments, as merchant of record | Billing details | USA |
Website fonts and the charting library are served from our own server, so no third party receives your IP address when they load. The product analytics script is served from our own server too, and is loaded only after you accept analytics cookies; only then does your browser send data to our analytics service provider.
We will give account holders at least 30 days' notice by email, and update this section, before adding or replacing a sub-processor that handles trace data. You can object by contacting us; if we can't resolve the objection, you may delete your account. We will not send your trace data to any other AI provider without this notice.
5. AI processing
- When Rootline detects a failed run, it sends a third-party AI provider a short summary of that run to get an explanation. This happens during the sync that picks up the run.
- What is sent: the failing step's input, output and error, plus the run's user input and final answer, about 2,000 characters in total; also the failure type and a short detector note, the failing step's name and type, the tool name, the number of steps, and the run's latency and cost.
- Redaction: before sending, we automatically replace text matching these patterns: email addresses; Langfuse keys (
sk-lf-…,pk-lf-…); other common API-key formats (sk-…,sk-proj-…,sk-ant-…,re_…,sk_live_,rk_test_etc.) and webhook secrets (whsec_…); Rootline ingest keys (rl_…); AWS access key IDs (AKIA…/ASIA…); GitHub tokens (ghp_etc.); Slack tokens (xox…); Google API keys (AIza…); JWTs; bearer tokens; PEM private keys; any opaque string of 40+ letters/digits; and values after labels such asapi_key,key,secret,token,passwordorauthorization. This is pattern-based. It can miss unusual formats, and it does not remove other personal data such as names, phone numbers or addresses. Mask such data upstream (section 2). - What is not sent: full traces, other steps' inputs and outputs, your account details.
- We use our AI provider's API. Under the provider's API data policy, data sent through the API is not used to train its models unless the customer opts in. The provider may keep API inputs and outputs in abuse-monitoring logs for up to 30 days (longer only where required by law). We do not have a Zero Data Retention agreement with the AI provider; if we obtain one, we will update this section.
- Explanations are stored in our database and reused for repeat failures. AI-written explanations are labelled "AI" and may be wrong. If AI is unavailable, turned off, or a usage limit is reached, Rootline shows its rule-based explanation instead.
- You can turn AI explanations off for your workspace in Settings → AI explanations. When off, nothing from your traces is sent to any AI provider and Rootline shows rule-based explanations only. Explanations already written stay stored (not shown) until the 90-day limit in section 8 removes them, or until you delete your account.
6. Security
- Connections to Rootline use HTTPS; session cookies are HttpOnly, SameSite=Lax and Secure in production.
- Passwords are hashed with Argon2. Langfuse secret keys are encrypted in our database (Fernet) with a key kept outside the database. Ingest keys, reset tokens and verification codes are stored hashed.
- Data at rest: our database is a managed Postgres database, which our hosting provider states is encrypted at rest with AES-256.
- Each workspace's data is kept separate. Login and reset attempts are rate-limited.
- Admin access: a small number of named administrators (currently the operator) can view account details and a user's runs, issues and costs, and can change plans or disable accounts. Admins view trace content only when needed to handle a support request you raise, abuse, or a security incident. Every admin view of a user's account or data, and every admin action, is recorded in an audit log.
- No system is perfectly secure. We do not currently hold security certifications such as SOC 2 or ISO 27001.
- Breach notification: if we become aware of a breach affecting your data, we will notify affected account holders by email without undue delay and in any case within 72 hours of becoming aware, with what we know, what data was affected and what we are doing. We will also notify regulators where the law requires.
- Reporting a vulnerability: email support@rootline.site. We will acknowledge within 3 business days.
7. Who runs Rootline
Rootline is operated by Ali Habib, trading as Rootline (Elegiac AI), Qatar. Elegiac AI is not yet a registered company; we will update this section when it is.
The operator is the controller of your account data and your processor (service provider) for the trace data you send or connect, as described in section 1. You can reach the operator at support@rootline.site.
8. Retention and deletion
- Trace data (synced Langfuse observations, OpenTelemetry spans and runs), AI explanations and alert records are deleted automatically once they are more than 90 days old (by the run's or record's timestamp). This check runs once a day.
- Account deletion: deleting your account in Settings (by typing your email) immediately and permanently deletes from our database your profile, sessions, codes, workspace, settings (including stored Langfuse keys), ingest key, and all synced Langfuse observations, OpenTelemetry spans, runs, AI explanations and alert records. Your data in Langfuse itself is not affected.
- Admin audit log: kept for 12 months, then deleted. When an account is deleted, its email address in the audit log is replaced with a one-way hash label ("deleted user #…").
- Backups: our database currently runs on a hosting plan that has no backups (our hosting provider offers no point-in-time recovery or backup exports on that plan), so deleted data is not kept in any backup. If we move to a paid plan, the provider keeps point-in-time recovery data for 3–7 days; we will update this section then.
- Disconnecting Langfuse deletes the Langfuse observations we stored for that workspace.
- Deletion on request: you can ask us to delete specific data by emailing support@rootline.site; we will do it within 30 days and confirm by email.
- Verification and reset codes expire automatically. Sessions expire after 14 days.
- Third parties: our AI provider may keep data sent for explanations for up to 30 days (section 5); our email provider keeps delivery logs under its own policy. Emails already sent stay in your inbox.
9. Your rights
Depending on where you live, you may have rights over your personal data. We will honour the requests below for all users where possible, and respond within 30 days (45 days for California requests where the CCPA allows).
- UK and EU (UK GDPR / GDPR): access, correction, deletion, restriction, objection, data portability, and withdrawing consent. You can complain to your data protection authority (in the UK, the ICO).
- California (CCPA/CPRA), where it applies: to know, delete and correct personal information, and not be discriminated against for using these rights. We do not sell or share personal information as the CCPA defines it.
- Canada (PIPEDA): access and correction, and to complain to the Office of the Privacy Commissioner of Canada.
- Australia (Privacy Act 1988): access and correction, and to complain to the OAIC.
How to make a request: you can update your name, email and password, turn AI explanations off, and delete your account yourself in Settings. Export: you can request a copy of your account data and stored trace-derived data by emailing support@rootline.site. For anything else, email us (section 14). We may need to verify your identity.
Your end users' data: if your traces contain personal data about your own users, you decide what is sent to Rootline and we process it on your behalf. Please send your end users' requests to you; we will help you respond.
Data Processing Agreement: a DPA (including the EU Standard Contractual Clauses and the UK Addendum) is available on request at support@rootline.site.
10. International transfers
We are based in Qatar. Rootline currently stores data only in the USA; we do not yet offer an EU data region. Our providers process data in the USA and elsewhere, so your data is transferred outside your country, including from the UK, EU, Canada and Australia. For transfers from the UK/EU, we rely on Standard Contractual Clauses (and the UK Addendum) in our DPA and with our providers, and on providers' EU-US Data Privacy Framework participation where available.
11. Cookies
We use strictly necessary cookies, plus an analytics cookie only if you accept it:
| Cookie | Purpose | Lasts |
|---|---|---|
rl_session | Keeps you signed in | 14 days |
rl_csrf | Protects forms against cross-site request forgery | 24 hours |
Analytics cookie (name starts with ph_), only if you accept | Product analytics: recognises your browser across visits (section 2) | 1 year |
rl_vid, only if you accept | Our own first-party analytics: a random identifier so a returning visitor is counted once (section 2) | 1 year |
rl_sid, only if you accept | Our own first-party analytics: groups the pages you view into one visit; renewed while you browse | 30 minutes |
rl_internal, Rootline team only | Set only when a member of our team opens a special link, so their own visits are left out of our statistics | 2 years |
We do not use advertising cookies. Our own first-party product analytics works without cookies (section 2); it uses rl_vid and rl_sid only if you accept analytics cookies. Analytics cookies are set only after you click Accept on our cookie banner; if you decline, they are never set. We remember your choice in your browser's local storage (rl_analytics_consent) so we don't ask again on every page, and we keep a record of each choice on our own servers (section 2): accept or decline, the time and the page, plus the random visit identifier only if you accept. To change your choice, use “Cookie settings” at the bottom of our public pages; declining after accepting stops session recording and deletes the analytics cookies.
12. Children
Rootline is a developer tool and is not meant for anyone under 16. We do not knowingly collect data from children under 16; if you think we have, contact us and we will delete it.
13. Changes to this policy
We will post updates here with a new "Last updated" date. For significant changes, we will email account holders before they take effect.
14. Contact
Privacy requests, security reports and questions: support@rootline.site. We aim to reply within 3 business days. Operator details are in section 7.