This policy explains what personal data KwikQA ("we") collects when you use the KwikQA command-line tool, website, and hosted service, what reaches the AI providers that produce diagnoses and repairs, and what you can do about it. It describes what the software and service do today; when that changes, this policy changes first.
In short
- KwikQA runs on your machine. Only a redacted, size-limited evidence packet that you have previewed and approved leaves it.
- On hosted plans, that packet passes through our service to the AI model and is never stored by us. The model's answer is held in memory until your CLI collects it, for at most 15 minutes.
- Every hosted request goes only to an AI endpoint with a zero-data-retention policy, so the AI providers don't keep it either.
- For a personal account, we store your GitHub id and username, and no email address. For a company account, we store the work email your company invited. Either way, we store your plan and metadata about each request, not its content, and we use no analytics or advertising cookies.
- In CI, we keep repositories and pull-request authors only as keyed hashes, for the month they count in.
- You can download your data or delete your account from your dashboard.
Who we are
KwikQA is responsible for the personal data described here: the data fiduciary under India's Digital Personal Data Protection Act 2023, and the controller under the GDPR. Contact us about privacy, or with a grievance, at privacy@kwikqa.com.
What we collect
Personal accounts
Hobby, Pro, and BYOK accounts sign in with GitHub. When you sign in with GitHub, we ask only for read access to your profile (the read:user scope). We store your GitHub numeric id and username. We don't store your email address or the GitHub token, and we get no access to your repositories. Your browser loads your profile picture directly from GitHub.
Company accounts
On the Teams, BYOK Teams, Enterprise, and BYOK Enterprise plans, accounts belong to a company account, whose admins invite people by work email. For an invitation, we store the address, the role and team it offers, and who sent it, and we email it. You sign in by entering a 6-digit code we email to that address. We store your work email with your account, and a keyed hash of it (HMAC-SHA-256, under a secret only our service holds) to find your account when you sign in. A code is stored only as a keyed hash, works for 10 minutes, and an address gets at most 5 codes an hour. Starting a company account works the same way, and we also store the company's name. To stop abuse, Cloudflare counts code requests and attempts from each IP address, and invitations sent by each person, for one minute; we don't store the address.
The company manages the accounts of its people. Its admins see each person's work email, role, teams, when they joined, and how many analyses and fixes they used this month. A team's managers see their team's people with what each used for the team, and the work emails of the company's people, to add them to the team. Admins can suspend someone, which signs them out everywhere, and a suspended account is erased after 30 days unless an admin restores it. When a team reaches 80% of a monthly cap its admins set, we email its managers, or the company's admins.
The CLI signs in at kwikqa.com/device: npx kwikqa login shows a code, and you approve it there while signed in to the website. Until the CLI collects its token, at most 15 minutes, we keep a hash of the CLI's device code, a keyed hash of the code you approve, which account approved it, and the team it counts for. To stop abuse, Cloudflare counts sign-ins started from each IP address, and codes tried by each account, for one minute; we don't store the address.
Plan and billing
We store your plan, billing period, seats, the identifiers Paddle gives your customer record and subscription, and a record of each refund. When you cancel, we look up your earlier refunds at Paddle to apply the limit of one full refund in 12 months, without storing what we read. Paddle is our merchant of record: it collects your name, email address, billing address, and payment details to sell you the subscription and handle tax, under its own privacy policy. We never receive your card details.
For our own accounts, when Paddle tells us a payment, refund, or chargeback went through, we record its date and amounts (before tax, the tax, Paddle's fee, and what reaches us), and a hash of Paddle's identifier for it so it's recorded once. Nothing in that record says whose it was.
Organizations and teams
Every account belongs to an organization: a personal one, or its company's. We store the organization's name and plan, its people and their roles, its teams with their managers, members, and monthly caps, and which team each token and request counts for.
Insights
Only if an organization's admins turn on insights (for a personal account, you): for each analyze, fix, and validate run in that organization, we store the titles, workspace-relative file paths, and projects of failing and flaky tests, the test framework (such as Playwright, Cypress, WebdriverIO, JUnit, TestNG, or pytest), the outcome and reason code of each fix, how many files and lines it changed, a hash linking a fix to its validation, and who ran it. Everyone in the organization can see these on the Insights page. Insights never include evidence, source code, diffs, or error messages, and --local runs share nothing.
Result counts
From kwikqa 0.6.0, after each analyze, fix, and validate run and each fix pull request, the CLI reports the result as codes and counts: the stage (analysis, fix, validation, a fix pull request opened, or a fixed test failing again), its outcome and reason codes, the test framework, the CI system, and the kind of failure: the symptom, the probable owner, the cause, how strong the evidence was, and the kind of file a fix changed (a spec, page object, fixture, test data, helper, or config); and counts: failure groups, failing and flaky tests, files and lines changed. From kwikqa 0.7.0, an analysis or a fix also names the usage sessions and retry tickets behind its model requests, as SHA-256 hashes that the service already keeps with each request, so we can count our model cost by kind of failure; the tickets themselves are never sent. It never reports test names, file paths, error messages, evidence, source code, or the model's own words. We store each result with the organization it counts for, never the person, so we can see which kinds of failure KwikQA diagnoses and repairs well, and at what model cost, and sum them into monthly totals. It's on by default: you can turn it off on your Account page, and a company's admins for everyone in it; --local runs, and runs with KWIKQA_RESULTS=off, report nothing.
CI runs
When KwikQA runs in a CI job, each metered request from that job also says where the run happens: the CI system (GitHub Actions, Azure Pipelines, GitLab CI/CD, or Jenkins), the host of the repository, its owner and name, its visibility when the CI system gives it, and the login of the pull request's author. On GitLab, the author is the user who triggered a merge-request pipeline. On Azure Pipelines, it is the id of the person who requested a pull-request build, never their email. A run without a pull request sends no author. On GitHub Actions and GitLab.com, the request can also carry the job's identity token, which the CI system signs and which names the repository and the person who triggered the run; we check its signature and take the repository and that person from it, then discard it. We store the repository and the author only as keyed hashes (HMAC-SHA-256, under a secret only our service holds), never the names themselves. The hashes are kept for the month they count in, toward your plan's repository or seat count. Our daily cleanup removes them after that month ends, and deleting your account removes the hashes billed to it.
Single sign-on and directory sync
Only for Enterprise company accounts that set them up: the company's connection (the SSOJet tenant and connection identifiers, the email domains it verified, whether it's paused, whether its directory syncs, and whether admins may use emergency codes). When you sign in through it, your identity provider tells us your work email and an identifier for you there; we keep the work email with your account, as for any company account, and the identifier only as a keyed hash (HMAC-SHA-256, under a secret only our service holds), with when you last signed in through it. From the company's directory, a keyed hash of each email address it provisions, with its domain and whether it's active, and the identifiers of the directory events we applied. We never store the identity provider's tokens or the ID token. A sign-in through single sign-on can create your account in the company, as a member.
Company audit log
For every company account, a record of who was invited, joined, suspended, or restored, changes to roles and teams, CI tokens created or revoked, renames, changes to its plan and seats, and downloads of the company's data. Each entry names who acted (a person, billing, or our operators) and what it concerns: a person, by work email, a team, or a token. When a person's account is erased, the log keeps their work email for the rest of its 400 days. Admins of Enterprise and BYOK Enterprise accounts can read their log.
Requests
For each AI request made through the hosted service, we store when it happened, its kind (diagnosis or repair), the model and provider, the test framework of the run it was about, its HTTP status, whether it counted toward a plan and whose, token counts, its cost to us, and a SHA-256 hash of the request. On bring-your-own-key plans, the CLI reports each request's kind, model, provider, and test framework, so it counts toward your plan. We also keep monthly usage counters, one-use retry tickets, and analysis sessions, which record how many requests one analysis or fix may make.
Each day, we sum each month's requests into monthly totals, which we keep longer than the requests themselves: for each organization, how many requests of each kind it made, how many counted, how many failed, and their cost to us; and for the whole service, the same by kind, test framework, and plan, which name no organization or person. A personal account's data export includes its totals.
Sharded CI runs
Only on Enterprise plans, when the shard jobs of one CI run each run KwikQA: which job claimed each failure group, so the run's other jobs don't diagnose it again. We store keyed hashes of the run, the job, and the group's fingerprint, never the run's identifier, repository, job, tests, or errors, and delete them after 24 hours.
Model jobs
While a hosted model request runs, the service keeps a job record of the account, the request's id, and whether it has been answered, so it can give back a request that went unanswered. The job, and the answer it holds, are deleted when your CLI collects the answer, and in any case after 15 minutes. The answer is held only in memory and is never written to storage.
Tokens and sessions
We store SHA-256 hashes of your CLI and CI tokens and website sessions, never the tokens themselves, along with the names you give tokens, when each was created and last used, the last KwikQA CLI version it was used with, which the CLI reports from version 0.5.0, and when it was last used by a CLI that didn't report one.
Cookies
The website sets four cookies, all strictly necessary for signing in:
| Cookie | Purpose | Lasts |
|---|---|---|
__Host-kwikqa_session | Keeps you signed in. Pages' scripts can't read it. | 30 days |
__Host-kwikqa_signed_in | Tells the site's pages that you are signed in. It holds no identifier. | 30 days |
__Host-kwikqa_oauth | Protects the GitHub sign-in step. | 10 minutes |
__Host-kwikqa_sso | Protects a single sign-on step: it ties the sign-in to your browser and the company you're signing in to. | 10 minutes |
We use no analytics, advertising, or tracking cookies, and no third-party scripts except Paddle's checkout, on the checkout page only.
Service logs
Our hosting provider, Cloudflare, processes IP addresses and request details to deliver the service and protect it from attacks. We don't store IP addresses. When a request fails unexpectedly, our logs record only its method, path, and the kind of error, never its content.
Support
If you email us, or reply to an email the service sent you, we receive your email address and what you write in our mailbox at Zoho Mail, and keep it for as long as we need to help you and keep a record of the request. Please don't send evidence, secrets, or source code by email.
What we don't collect
- We never store your source code, evidence packets, diffs, or the model's answers.
- On bring-your-own-key plans, evidence goes from your machine directly to the model endpoint you choose and never reaches our servers.
- Run records, patches, transaction journals, captures, and validation output stay on your machine, in
.kwikqa/in your project. - The CLI sends no telemetry or analytics. It contacts our service only for your account (sign-in, plan, and billing), to count and send the requests you approve, to say where a CI run happens, and to share insights where an organization's admins turned them on.
What reaches AI providers
The CLI builds an evidence packet on your machine from your failing tests: selected report fields, trace events or the Cypress and WebdriverIO command log, the page elements around each failed locator, the accessibility snapshot of the page, failed requests and console errors, short source excerpts, and recent code changes that mention the failure, with their commit subjects but never author names or emails. Screenshots and videos are described by size only. Before you see the preview, the CLI removes the credentials and secrets it recognizes: password and token fields, authorization and cookie headers, credentials in URLs, and values named like passwords, secrets, tokens, or API keys.
Redaction recognizes credentials, not every kind of personal data. Names, email addresses, or other personal data in your test data or on the pages under test can appear in the packet. Use test data you may share, check the preview, and decline a request you don't want to send. Security describes the packet in detail.
When you approve a request on a hosted plan, our service forwards the packet unchanged to OpenRouter, which sends it to the model endpoint named in the preview:
- Claude, for diagnoses and repairs, on Google Cloud Vertex AI; and
- GPT, for a second repair attempt after the first fails validation, on Microsoft Azure.
Every hosted request asks OpenRouter for zero data retention. Our service refuses any request that doesn't, and OpenRouter sends it only to an endpoint whose policy is to keep neither the request nor the answer, or refuses it. OpenRouter itself doesn't store the content of requests; it keeps metadata such as token counts and timing. Because the request isn't kept, it isn't used to train models. These are the providers' own commitments, under their terms with OpenRouter: we check on every request that zero retention was asked for, but we can't inspect their systems.
On bring-your-own-key plans, requests go from your machine to the model endpoint you choose (OpenRouter, the Anthropic API, the OpenAI API, Azure OpenAI, Amazon Bedrock, or GitHub Models), under your own account with it, its privacy settings, and its terms. KwikQA's default OpenRouter routes ask for zero data retention there too; a route you choose yourself doesn't, and on the other endpoints, whether a request is kept is up to your account with that provider.
How we use it
- to provide the service: sign-in, plans, usage limits, organizations and teams, the emails it sends, and forwarding the requests you approve;
- to bill you, refund you, and keep billing records;
- to keep the service secure and prevent abuse;
- to answer support requests; and
- to monitor the service's costs and usage in aggregate.
We don't sell or rent personal data, use it for advertising, or use your data to train AI models.
Why we may process it
Under India's Digital Personal Data Protection Act 2023, we process the personal data you give us for the purpose you gave it, which is to provide the service you signed up for, and with your consent where that is needed. You can withdraw consent at any time by deleting your account; that doesn't affect what was done before.
Where the GDPR or UK GDPR applies, we process account, request, and billing data to perform our contract with you, security and cost-monitoring data for our legitimate interest in running a reliable and affordable service, and billing records to meet legal obligations.
Who processes it
| Provider | Purpose | Data |
|---|---|---|
| Cloudflare | Hosting, database, network protection, and sending email (sign-in codes, invitations, team-cap notices, notices that an admin signed in with an emergency code, and service alerts to our own operators, which name no customer), and the service's logs, which hold codes, counts, and ids, never tokens, request contents, email addresses, or IP addresses | Account, plan, usage, and request records; work email addresses and the emails sent to them; IP addresses, in transit |
| OpenRouter | Routing approved requests to the AI model | Evidence packets, in transit; request metadata |
| Google Cloud Vertex AI | Claude diagnoses and repairs, with zero data retention | Evidence packets, in transit |
| Microsoft Azure | GPT repair retries, with zero data retention | Evidence packets, in transit |
| Paddle | Merchant of record: payments, tax, invoices, refunds, and the customer portal | Billing details |
| GitHub | Sign-in and profile pictures | GitHub identity |
| SSOJet | Single sign-on and directory sync for Enterprise company accounts, between our service and the company's identity provider | People's sign-in identity (name and work email) and directory records, from the company's identity provider |
| Zoho Mail | Our support mailbox (support@kwikqa.com), which also receives replies to the emails the service sends | Your email address and what you write to us |
We'll update this list before a new provider receives your evidence. We also disclose data when the law requires it, and to a successor if the business is transferred, under this policy.
Where it is processed
Our database's primary copy is in Cloudflare's Asia-Pacific region, and Cloudflare serves the website and service from the data center nearest you. Our support mailbox is in Zoho's India data center. OpenRouter, the AI providers, Paddle, GitHub, and SSOJet (whose region for our service is the United States) may process data in the United States, the European Union, and other countries. India's law allows these transfers except to countries the Government of India restricts. Where the GDPR or UK GDPR requires it, transfers rely on safeguards such as the European Commission's standard contractual clauses.
How long we keep it
| Data | Kept |
|---|---|
| Account and plan | Until you delete your account, or for a company account, until the company deletes it or its account |
| Suspended people in a company account | 30 days, then erased |
| Company accounts never paid for | 7 days |
| Request records and monthly usage counters | 13 months |
| Monthly totals of requests | 5 years; an organization's go when its account is deleted |
| Payment, refund, and chargeback amounts, without whose they were | 8 years |
| CI repository and author hashes | The month they count in; removed by the daily cleanup after it ends, or when the account is deleted |
| Model jobs, and the answers they hold in memory | Until your CLI collects the answer; at most 15 minutes |
| Analysis sessions | 90 days after they close |
| Retry tickets | 90 days |
| Website sessions | 30 days |
| CLI sign-in codes | Until the CLI collects its token; at most 15 minutes |
| Work-email sign-in codes (keyed hashes) | Valid for 10 minutes; removed by the daily cleanup after they expire |
| Active tokens | Until revoked, or until they expire: 90 days without use, or sooner when a company caps token lifetimes |
| Revoked and expired tokens | 90 days after revocation or expiry |
| Invitations | Until accepted or revoked, or the company account is deleted |
| Team memberships | Until the person is removed from the team, or the team or account is deleted |
| Insights | 13 months, or until the organization's admins delete them |
| Result counts | 13 months, then as monthly totals for 5 years; an organization's go when its account is deleted |
| Company audit log | 400 days, including the work email of people erased in that time |
| Single sign-on identities | Until the person's account is erased, the company's connection is removed or moves to another tenant, or the company account is deleted |
| Directory records from the company's identity provider | Until the company's connection is removed or moves to another tenant, or the company account is deleted |
| Identifiers of applied directory events | 30 days |
| Shard jobs' claims on a CI run's failure groups | 24 hours |
| Refund records | Until the account is deleted |
| Evidence, source code, model answers | Never stored |
| Billing records held by Paddle | As long as tax and accounting law requires |
Your choices and rights
- Access and export: download your account, organization role, teams, tokens, requests, and insights, and for a personal account its usage, monthly totals, and result counts, as a JSON file from Account in your dashboard; a company's admins can download the company account's records from its Organization page. Email us for anything else we hold.
- Erasure: delete a personal account from the same page. This cancels any subscription, refunding as cancelling would under the refund policy, and at once removes your account, tokens, sessions, requests, usage, audit events you caused, insights, refund records, and CI repository and author hashes. In a company account, the company decides about your account: ask its admins to remove you, or email us. Its admins can delete the whole company account, which removes its people, teams, tokens, usage, audit log, and everything else it holds at once. Paddle keeps the billing records the law requires.
- Correction: your GitHub username updates each time you sign in. Email us to correct anything else.
- Nomination: under India's law, you may nominate someone to exercise these rights for you if you die or become unable to. Email us to do so.
- Other rights: depending on where you live, you may also object to or restrict processing.
Email privacy@kwikqa.com for any of these, or with a grievance about how we handle your data. We reply within 30 days. If you aren't satisfied, you can complain to the Data Protection Board of India, or to the data protection authority where you live.
Security and breaches
Data is encrypted in transit and at rest, tokens and sessions are stored only as hashes, and access to production systems is limited to KwikQA's operator; Security has the details. If a personal data breach affects you, we'll tell you, and the authorities the law requires, including the Data Protection Board of India, without undue delay and within the time the law sets.
Children
KwikQA is for professional use and is not intended for anyone under 18. We don't knowingly collect children's data; if you believe a child has an account, email us and we'll delete it.
Changes
We'll post any changes here and update the effective date. For material changes, such as a new kind of data or a new provider that receives evidence, we'll tell you in the dashboard before they apply.