Legal
Privacy Policy
Effective October 27, 2026
This policy explains what DeployAngel collects, why, how long we keep it, and the choices you have. DeployAngel is run by Jordan Owens, an individual based in Ohio, USA ("we" and "us"). It covers the website at deployangel.com, the dashboard and API at app.deployangel.com and api.deployangel.com, and the deployangel agent.
The short version
- The agent sends summaries of what your app does, not your users' data: no request bodies, cookies, headers, parameters, or SQL values.
- We use your data only to run DeployAngel. We don't sell it, and we don't use it for advertising.
- AI investigation is off by default. When it's on, a small evidence bundle goes to Anthropic for each failed release.
What we collect
Account information
- Your name, email address, team name, and role.
- Your password, stored only as a secure hash.
- Whether you've verified your email address, and which of each account's emails you get.
- The email addresses of people you invite.
- Where you came from when you signed up: the tag on the link you followed (such as
?ref=hn), or else the name of the site that linked to us, such as news.ycombinator.com. We use it only to learn which places bring sign-ups. - For each sign-in session: your IP address, your browser's user agent, and when the session started.
- If you sign in through Heroku: your Heroku email address and user ID.
- Your answers when you tell us whether a release's verdict was right or its AI investigation was useful, and any note you add. They're kept with the release. When you say a verdict was wrong or an investigation wasn't useful, your answer is also emailed to us.
- If you set up report branding, the name and logo you add.
Telemetry from the agent
The agent summarizes what your app does inside your app, and sends that summary about once a minute:
- Request counts, status codes, and response-time histograms for each normalized route, such as
GET /users/:id. IDs in paths are never sent. - Background job counts, failures, durations, and queue delays for each job class.
- Exceptions: the exception class, where in your code it happened, and its first line of message, with numbers, IDs, email addresses, quoted text, the request's host, and (with ros-apartment) tenant names replaced by placeholders. Other words in a message are kept, so a message your app builds with a customer's name in it would include that name. You can turn messages off in the agent, and then none is sent. Occasionally, a stack trace limited to your application's own files.
- The release version and commit, and which process sent it, such as the dyno or host name.
- The names and counts of any checkpoints and critical flows you define.
- Unless you turn it off with
DEPLOYANGEL_FILE_DIGESTS=false, the paths and SHA-256 hashes of your app's files, so we can tell which files changed between releases. Never their contents.
Once when each process starts, the agent also sends a description of your app's code, so DeployAngel knows what a release should exercise and which files a failure points to:
- The route table: each route's pattern and HTTP method, the controller or view that serves it, and, where the agent can tell, that code's file path.
- The names of your background job classes or tasks, and, where the agent can tell, their file paths.
- The recurring jobs your app declares: each one's name, the job it runs, its schedule, and its time zone.
- The runtime: the language and framework, and their versions.
These describe your code, not your users or their data.
The agent never sends request or response bodies, headers, cookies, query parameters, session data, SQL parameter values, or records from your database.
Releases and connected services
- Deployments: the version, commit, time, and where each came from, such as a Heroku webhook, your CI, or the agent's telemetry, when DeployAngel notices a new release itself.
- GitHub, if you connect it: access to read the repository you choose, used to work out what changed in each release, to add a check with each release's verdict to its commit, and, when a release fails, to comment on the pull requests it shipped with its verdict summary. Anyone who can see the repository can see those checks and comments, so on a public repository they're public. We keep a summary of the changed files, dependency and framework versions, migrations, and pull request numbers, titles, and links, not your source code. The one exception is AI investigation, below.
- Heroku, if you use the add-on: your app and add-on names, your plan, and release events.
- Slack: the incoming webhook URL you add, stored encrypted.
- Honeybadger and AppSignal: the API key you add, stored encrypted.
- PagerDuty: the integration key you add, stored encrypted.
- Webhooks: the URLs you add and their signing secrets, stored encrypted.
AI investigation (optional)
AI investigation is off by default for every app. When it's on and a release fails, we send Anthropic a small bundle of evidence: the failing findings, new exceptions, what changed, and, if GitHub is connected, short excerpts of the few most relevant changed files. The explanation that comes back is stored with the release, along with the evidence we sent. AI ethical use below covers how we use AI.
The contact form
Your name, email address, and message, emailed to us so we can reply. If you're signed in, the email also says which account you're in, and the dashboard page you opened the form from. DeployAngel itself doesn't keep a copy.
Cookies, fonts, and logs
- The dashboard uses essential cookies to keep you signed in and to protect its forms. There are no analytics, advertising, or tracking cookies.
- deployangel.com sets no cookies. It runs two scripts. Cloudflare Web Analytics counts page views, which sites visitors came from, and how fast pages load, without cookies and without identifying you; the dashboard doesn't use it. Our own small script passes the link's tag, or the name of the site that linked to us, along to the sign-up form by adding it to the page's links. It stores nothing in your browser.
- Both sites load fonts from Google Fonts, so Google receives your IP address and browser details when a page loads.
- We and our hosting providers keep standard request logs, such as IP addresses and times, for security and reliability.
How we use it
- To run DeployAngel: verify your releases, show them in the dashboard, send the notifications you set up, and answer your messages.
- To keep the service secure, prevent abuse, and fix problems.
- To tell you about important changes to the service or to these policies.
We don't sell personal data, and we don't use it for advertising.
AI ethical use
This section is DeployAngel's AI ethical use policy. AI investigation is the only part of DeployAngel that uses AI, and we hold it to these commitments.
What the AI does
When a release fails, AI investigation reads the evidence DeployAngel has already gathered and writes an explanation: the most likely cause, the evidence behind it, the files to look at, a next debugging step, and what's still uncertain. It runs on its own for failed releases on apps where it's turned on, and when someone on your team asks for one on a release's page.
What it never does
- Decide verdicts. Whether a release is cleared, failed, or not cleared yet is decided by fixed rules that don't use AI. The AI is told not to second-guess the verdict.
- Act on your app. It can't deploy, roll back, or change anything in production or in DeployAngel. It's also told not to recommend a rollback: that decision belongs to your team.
- Make decisions about people. It isn't used to evaluate, rank, or profile developers, your team, or your app's users.
People stay in charge
Each explanation is labeled as AI inference, not a verdict. It comes with a confidence level (high, medium, or low), the specific findings, exceptions, and files it relies on, and its caveats. When the evidence is thin, it's told to say so. AI can be wrong, so treat an explanation as a lead to check, not a conclusion.
Your choice
AI investigation is off by default for every app. Anyone on your team can turn it on or off for an app in that app's Settings. Each account also has a daily limit on investigations.
Your data
- The AI sees only the evidence bundle described in AI investigation above. It's built from telemetry that leaves out your users' data.
- Investigations use Anthropic's Claude models through Anthropic's commercial API. Under Anthropic's commercial terms, Anthropic doesn't train its models on that data.
- We don't use your data to train AI models.
- The evidence we sent and the explanation are kept with the release, and deleted along with it.
Changes and concerns
Before we use AI anywhere else in DeployAngel, or switch AI providers, we'll update this section and the provider list below. If an explanation is wrong in a way that misled you, or anything about it concerns you, tell us through the contact form.
Who else processes it
We use these providers to run DeployAngel. Each gets only what it needs for its job:
- DigitalOcean hosts the dashboard, API, and database.
- Heroku (Salesforce) bills customers who use the Heroku add-on.
- Cloudflare hosts deployangel.com, serves our DNS, and counts the site's page views.
- Resend sends our emails: notifications, invitations, email address verification, password resets, notices when your email address or password changes, and contact form messages.
- Honeybadger receives error reports when part of DeployAngel fails. A report can include the request that failed, your IP address and browser, and your user ID. Passwords, tokens, and the cookie that keeps you signed in are removed first.
- Anthropic runs AI investigation, only for apps where it's turned on.
- Google Fonts serves the fonts on both sites.
GitHub, Slack, Honeybadger, AppSignal, PagerDuty, and your webhooks only get requests when you connect them: we read from GitHub, add a check with each release's verdict to its commit, and comment a failed release's verdict summary on the pull requests it shipped, post to the Slack webhook you add, post release verdicts, including their findings and exception summaries, to the webhook URLs you add, and send each verdict's one-line summary, release, commit, and a link to it to the Honeybadger project or AppSignal app you add, and send release verdicts, with the same summary and details Slack gets, to the PagerDuty service you add. If you share a release, anyone with its link can see that release's page: its verdict, findings, coverage, and history, and what changed only if you choose to include it. Stopping sharing turns the link off. We may also share data when the law requires it, or to protect the service or other people from harm. If DeployAngel is ever sold or merged, your data would move with it, still covered by this policy.
How long we keep it
- Release history: 7, 30, or 90 days, depending on your plan. Each app's latest release and its last two cleared releases are kept as baselines.
- Daily totals of each app's requests, server errors, exceptions, and job runs, used for monthly reports: as long as your plan's release history.
- Telemetry: 21 days on every plan, so baselines have enough history to learn from. After a day, per-minute data is combined into hourly totals.
- The description of your app's code (its route table, job names, schedules, and file paths and hashes): as long as the release it describes. The most recent one is kept until a newer one replaces it.
- Account information: until you close your account.
- Deleted apps: deleted right away, with everything collected for them. Apps whose Heroku add-on is removed are deleted 7 days later.
Deleted data can remain in our database provider's encrypted backups for up to 7 days, until those backups expire.
How we protect it
- All traffic uses HTTPS.
- Passwords are hashed. API tokens and invitation links are stored only as hashes, so we can't read them back.
- Slack and webhook URLs, webhook signing secrets, Honeybadger and AppSignal API keys, PagerDuty integration keys, and Heroku secrets are encrypted in the database.
- Each account's data is kept separate, and only its members can see it.
Your choices and rights
- Ask us to access, correct, export, or delete your personal data through the contact form. Account owners can also delete apps in Settings at any time.
- Turn off file digests in the agent, and leave AI investigation off.
- Choose which emails you get from each account in your profile, or unsubscribe with the link in any of them.
- Depending on where you live, for example in the EU, the UK, or California, you have further rights, such as objecting to processing, and complaining to your data protection authority. We'll respond to requests within 30 days.
- DeployAngel is built not to receive personal data about your app's users. If you think it has, tell us and we'll delete it.
Where your data is processed
DeployAngel is hosted in the United States. If you use it from somewhere else, your data is transferred there and processed under this policy.
Children
DeployAngel is for professionals, not children. We don't knowingly collect data from anyone under 16.
Changes to this policy
We'll post any changes here and update the date at the top. For material changes, we'll tell account owners by email or in the dashboard before they take effect.
Contact
For privacy questions or requests, use the contact form and choose "Privacy or security".