Privacy Policy

Last Updated
November 20, 2026
Sections
10

Latchkey is committed to protecting the privacy and security of your data. This Privacy Policy explains how we collect, use, and safeguard your information.

§01 Introduction

Latchkey ("we," "our," or "us") is committed to protecting the privacy and security of your data. This Privacy Policy explains how we collect, use, and safeguard your information when you use our website located at https://latchkey.dev/ (the "Site") and our AI-powered CI/CD monitoring services (the "Service").

By using our Service, you agree to the collection and use of information in accordance with this policy.

Effective Date: October 1, 2026
Contact: legal@latchkey.dev

§02 Information We Collect

We collect only the minimum amount of data required to run and heal your jobs, provide actionable insights and optimize your pipelines.

2.1 Account & Billing Information

DataDetailsRetention
AuthenticationWe use Clerk as our managed authentication provider. You sign in with one of the methods the sign-in page offers, which may include GitHub, Google, or a code sent to your email address. We do not collect, view, or store your passwords.§06
Google Sign-InIf you sign in with Google, Latchkey receives your basic Google profile through Clerk (Google's openid, email, and profile scopes), which includes your name, email address, and profile picture. We keep your name, your email address, and a link to your profile picture with your account, and nothing else from your Google account. We use them to identify you, to show your name and picture in the dashboard and to the members of your workspace, to bill your workspace, and to email you about your account and the Service. We do not sell this information or use it for advertising, and we share it only with the service providers that help us run the Service, such as Clerk for sign-in, Stripe for billing, and our email provider.§06
BillingAll payments are processed securely by our third-party payment processor, Stripe. We do not collect or store your card numbers, bank account numbers or other financial account numbers on our servers.§06

2.2 Pipeline & Usage Data

DataDetailsRetention
Workflow MetadataRun IDs, job names, timestamps, failure/success statuses, and duration metrics.§06
Workflow YAML FilesWe read your .github/workflows YAML files to analyze job structure, caching, and dependencies. This powers AI-driven optimization recommendations and AI Scan runner configurations. We also read them so that self-healing can propose fixes to them, to make security findings about the actions they use, to change their runs-on labels when you move jobs to Latchkey runners, to plan the Latchkey CI workflow files that replace them when you move CI to Latchkey, to tell build configuration from application code when we scan a repository, and to index them in your knowledge base. For Latchkey CI, we also read your workflow files under .latchkey/workflows, to run the jobs they define, to make AI-driven optimization recommendations about them, and so that self-healing can propose fixes to them.§06
Resource UsageCPU and memory usage statistics used to calculate cost efficiency.§06
Log DataBuild logs and error messages required to identify bottlenecks and failures.§06
Runner TelemetryWhen you use Latchkey Runners, we process runtime telemetry - CPU, memory, disk usage, exit codes, job durations, and per-runner metering data - to operate the runners, bill usage, surface analytics, and power self-healing. A runner created after our network change on November 3, 2026 reaches the internet through its own public address, not through our NAT gateways, and no flow log records which remote addresses its connections reach. What a step prints, including any host or address it names, is still handled as Log Data and Self-Healing Agent Data describe. For a runner created before that change, our NAT gateways record which remote addresses its outbound connections reached and how many bytes passed, in totals of up to ten minutes with no packet content, no ports and no process or user name. Only our operators read that record, to investigate network cost. It is kept for 14 days and deleted within three days after that, and an operator's copy of any part of it is deleted within 14 days of being made.§06
Self-Healing Agent DataWhen self-healing engages on a failing build step, the agent processes that step's stderr, stdout, and exit code, and can read files in the job's checkout or in the working tree you uploaded from the Latchkey CLI: at most four lookups for each diagnosis and the first 8 KiB of a file. What it reads goes to our AI model provider and is kept in the heal record for 90 days. Its actions are bounded to a vetted allowlist.§06
Pull Request Changed FilesWhen a pull request runs on Latchkey CI, we read from GitHub the list of files the pull request changes and keep their paths, with the pull request's number and commit, for 90 days, to check, when self-healing re-runs a test, that the test does not reach changed code, and to determine which of the pull request's jobs no change reaches. To verify the list, we also read the pull request itself. GitHub's answers can carry each file's changes and the pull request's title and description, and we keep none of them.§06
Job OutputsWhen a Latchkey CI workflow file names outputs for a job, we keep the values the job's steps write for those names once the job succeeds, at most 1 MiB for each job, so that the jobs that need it can read them, in that run and when you re-run it. We keep them until you delete the workspace. A value holding anything that job's log masks, such as a secret the job was given, is not kept; a secret a step fetches for itself is kept as written unless that job's log masks it (a step can hide a value with ::add-mask::).§06
Source Code: Where It Is KeptLatchkey keeps your source code where you or your jobs ask it to, and what self-healing and Latchkey's other features read of it, as this row describes. Analytics and AI insights read metadata, workflow YAML files and logs, and no other file of your repository. When you use Latchkey Runners, your code is checked out into an ephemeral, isolated runner instance solely to execute that job, and the runner is destroyed at job completion. After a job, your code stays on Latchkey infrastructure in a cache entry a job saves, an artifact a job uploads to Latchkey or an output a job records. Each is available only within your workspace and is deleted when the period our documentation states for it ends. It also stays in a heal record, which keeps what self-healing read of your files for 90 days, and in a fix self-healing proposes, whose file contents are kept for 365 days. A working tree you upload from the Latchkey CLI to start a job is deleted after that job ends, apart from what self-healing read of it, which the heal record keeps for those 90 days. Latchkey's other features keep parts of your code as well: the knowledge base keeps the docs, CODEOWNERS and other community files, and CI configuration it indexes while the repository is enabled, and afterwards for the periods section 6 states; our records of AI Scan and repository scan runs keep what those runs read, including whole workflow files and up to 8 KiB of each file they open, which can be any file of the repository, for 90 days; an AI Scan's original and optimized workflow files, the change an optimization recommendation proposes, and the action references a security finding records are kept until you delete the workspace; a Latchkey CI job's step commands, compiled from its workflow file, are kept for about a day; and a trial run keeps the first line of a failed step's command and the end of its output for 180 days.§06
  • AI Privacy: We do not use your proprietary logs or source code to train global, multi-tenant AI models. AI insights, self-healing diagnoses, and AI Scan results are generated for your workspace only.
  • Permissions Scope: For analytics, we read GitHub Actions metadata and workflow YAML. Our access to GitHub Actions also lets us read job logs, dispatch a workflow when you or your coding agent asks, and cancel a run on a repository that is no longer monitored once a job in it has waited 10 minutes for a Latchkey runner, the run's other jobs included, wherever they run. To open pull requests, we request write access to code, workflow files and pull requests, and we use it to write only to new branches of our own and to the pull requests we open from them, including the heal pull requests that self-healing opens on its own. We never push to your branches and never merge code. The same pull request access lets us read a pull request and the list of files it changes. For Latchkey Runners, your workflow YAML directs jobs to Latchkey runners via the runs-on label. In an organization, we use the self-hosted runners permission to create a runner group of our own, choose which of your repositories can use it, and register and remove Latchkey runners in it; to do so we read your organization's list of runners, and when you move jobs to Latchkey runners, we also read its self-hosted and GitHub-hosted runners to pick the matching Latchkey size. We do not store what we read of your organization's runners. On Latchkey CI, we also write checks on your commits and read the groups your merge queue makes, and when you compare a move to Latchkey CI side by side, we read the checks on the commits compared, which we do not store. When you sign up, connect your GitHub account or join a workspace, we check whether your own GitHub account is a member of its organization, without reading a list of members; for a pull request from a fork on Latchkey CI, we read whether its author, a member who opens its waiting run, or anyone who approves its commit can write to the repository; for an organization, we read its GitHub plan and this month's GitHub billing usage when the App connects, and keep only the plan and the Actions minutes; and when you disable a repository or delete the workspace, we try to delete any webhook Latchkey created on the repository. Our documentation lists the App's permissions and what each is for.

AI Data Usage

By default, Latchkey does not use your proprietary logs, workflow files, or source code to train global, multi-tenant AI models. AI features - insights, AI Scan runner configurations, and self-healing - operate on your workspace's data for your workspace's benefit. We reserve the right to use de-identified, aggregated metadata (such as job durations and failure-category counts) to improve our algorithms and service performance, in a form that cannot be traced back to you or your workspace.

§03 How We Use Your Information

We use the collected data for the following purposes:

3.1 Service Delivery

To generate analytics dashboards, cost reports, and optimization insights.

3.2 Runner Operation

To provision, operate, and tear down the ephemeral Latchkey Runner instances that execute your workflow jobs, and to meter their usage for billing.

3.3 Self-Healing

To detect, diagnose, and apply bounded fixes to failing build steps on Latchkey Runners, and to show you in the dashboard what self-healing did, on the run where each heal happened and in a summary across your workspace.

3.4 AI Scan & Optimization

To analyze your repository structure and workflow YAML files in order to generate AI Scan runner configurations and surface optimization recommendations you can choose to apply via a Latchkey-created pull request.

3.5 Security & Reliability

To detect technical issues, security risks, or abuse of the Service.

3.6 Billing Automation

To facilitate seamless plan transitions and automated billing via Stripe.

3.7 Latchkey CI Runs

To determine which of a pull request's jobs no change reaches.

§04 Data Sharing and Disclosure

We do not sell, trade, or rent your personal data to third parties.

4.1 Cloud Hosting & Runner Compute

AWS hosts our application and the compute that backs Latchkey Runners, and a managed database provider hosts our database. Customer data and runner instances are kept in isolated environments.

4.2 Authentication

Clerk handles authentication for every sign-in method we offer. We do not see or store your GitHub or Google credentials.

4.3 Payment Processing

Stripe facilitates your subscription, seat, and runner-minute payments.

4.4 AI Processing

We use third-party AI model providers to power Latchkey's AI features, such as self-healing diagnoses, optimization insights, AI Scan runner configurations and the knowledge base. When a feature uses one of these providers, we send it the data that feature needs for that request. We do not authorize these providers to train their models on your data.

§05 Data Security

We use industry-standard security measures to protect your data: TLS 1.2 or higher for data in transit and AES-256 for data at rest. We do not store passwords or payment information.

Latchkey runs on AWS, which is SOC 2 compliant. Customer data and Latchkey Runner instances are kept in isolated environments with strict access controls, and Latchkey Runners are ephemeral - each runner is destroyed after the job it serves completes, so nothing on a runner carries into another job.

§06 Data Retention

We retain data in an active status for up to one year after which it is transferred to cold storage. While historical analysis remains available, retrieval from cold storage may require additional time to load. You may request full data deletion at any time following the cancellation of your plan. Archived data is typically purged after five years unless otherwise required by law.

Where this Policy or our documentation states a period for a kind of data, such as a job's logs, a dependency cache entry, an artifact, a heal record or a proposed fix, that period applies instead: the data is deleted after the period ends and is not moved to cold storage. A record about that data, such as an artifact's name and size, or a summary built from it can stay longer.

Knowledge base documents are not moved to cold storage: they stay in the live knowledge base for as long as they are kept, as follows. Documents you upload to the knowledge base are kept until you delete them or your workspace. Documents indexed from a repository are kept while the repository is enabled and deleted when you disable it. A repository stops being available to us when it is removed from the GitHub App's access (which is also how we see a repository deleted on GitHub) or when the GitHub App is uninstalled; its documents are then deleted in the first weekly cleanup at least 90 days after the first such stop, and only giving the repository back to the GitHub App's access starts that period again: a repository given back but not enabled again counts it from when it was given back, or from the last change we recorded for it, such as a rename or the GitHub App being uninstalled, if that is later. A suspension of the GitHub App does not by itself start that period. A file that leaves the repository is deleted in the first weekly cleanup at least 90 days after we next index the repository and find it gone, or with the repository if that comes first. A document of a stopped repository, other than a file we found gone, is kept past that period while an enabled repository in the same workspace has the name it was indexed under, letter case included, until we hold that enabled repository's own copy of the file or index that repository in full after the stopped one's period began. Where our records cannot show that a repository stopped, its documents are kept until you disable it or delete your workspace. Documents indexed before we recorded which repository each document came from are kept until you delete them or your workspace, or until you disable a repository while it has the name they were indexed under, letter case included. A stop that happens while we are indexing the repository can delete some of its documents at once, and leave a few written in those minutes to count their period from when they were written.

§07 Your Privacy Rights

Depending on your location (e.g., GDPR, CCPA), you may have rights regarding access, correction, or deletion of your personal data.

§08 Cookies and Tracking Technologies

We use essential cookies for authentication and session management. Refer to general data collection practices outlined above.

8.1 Analytics

We use Google Analytics 4 to understand how the site is used: which pages are visited, which links and buttons are used, including the buttons that copy a setup command, which parts of a page are reached and how far down people read, the steps of first-run setup in the dashboard, and how quickly pages load. We do not send Google your name, email address, repository names, or the contents of any workflow. Signed-in visitors are described only in buckets, such as a plan tier, never by an identifier that points back to a person.

Analytics is on by default, so Google Analytics sets its cookies (_ga and _ga_*) when you visit. You can turn analytics off at any time under Cookie Settings in the site footer. Once it is off, nothing is sent to Google at all and any _ga cookies are removed from your device.

You can change this at any time using Cookie Settings in the site footer. Google processes this data in the United States on our behalf. We do not use Google Analytics for advertising, remarketing, or ad personalisation, and those categories stay switched off whatever you choose.

If analytics is on when you subscribe, we attach your analytics identifier to the checkout so the purchase can be counted in Google Analytics. Stripe receives that identifier with the checkout session; it is never sent to Google with your name or email.

§09 Third-Party Services

Refer to the Data Sharing and Disclosure section for information about our infrastructure providers, and to Cookies and Tracking Technologies for how we use Google Analytics.

§10 Contact Us

If you have questions about this Privacy Policy, please contact us at:

legal@latchkey.dev

Policy Updates

We may update this Privacy Policy from time to time. We will notify you of any material changes by email or through a prominent notice on our platform.

End of document · Privacy Policy