1. Who we are and what this policy covers
Xybrid Inc. (“Xybrid”, “we”, “us”) is a Delaware corporation at 251 Little Falls Drive, Wilmington, DE 19808, United States. For anything about privacy, write to [email protected].
This policy explains how we handle personal data when you use:
- our website, www.xybrid.ai;
- the console (console.xybrid.dev) and our sign-in service (auth.xybrid.dev);
- the Xybrid API (api.xybrid.dev), including access to it through MCP;
- trace and telemetry intake (ingest.xybrid.dev);
- our two model gateways (api.xybrid.dev and gateway.xybrid.dev);
- the model registry (registry.xybrid.dev);
- our documentation (docs.xybrid.dev).
We call these “the services”. The policy also covers the data the open-source Xybrid SDK and CLI send to the services. The SDK and CLI run on your own devices. When you use them to reach other websites from your machine, for example with the CLI’s web search or page fetch tools, those sites’ own policies apply.
This policy does not cover other apps we run at their own addresses, such as Sonix, Elixir or our investor site. It also does not govern data we process for our customers, which section 3 explains.
Our Terms of Service govern your use of the services.
2. Key points
- You don’t need an account to try Xybrid. A coding agent can create a free workspace for you. We collect very little at that point, and no name or email address.
- Traces are off until a developer turns them on. Once on, they can include prompts and model outputs unless the developer chooses a metadata-only setting.
- We delete unclaimed workspaces. If nobody claims a workspace by its claim deadline, we delete its account records and stored content. Traces it sent expire on the schedule in section 8.
- We don’t sell personal data and we don’t use it for cross-site advertising.
- Our website sets no cookies. Its analytics count page views without cookies.
- You can ask us to access, export, correct or delete your data by writing to [email protected]. We handle these requests by hand.
3. Data we control and data we process for customers
Data we control. We decide how to use your account details, website and registry analytics, signup records, logs and messages you send us. For this data we are the “controller” under European data protection law, and this policy applies to it.
Data we process for customers. Developers use Xybrid to collect traces and telemetry from their own apps, and to send model calls through our gateways. That data can include information about the developer’s own users, such as the prompts they type. For a claimed workspace, the organization that owns it decides what its apps send, and we process that content on its behalf, as its “processor” or “service provider”. If you use an app built with Xybrid and want to exercise your rights over that data, contact the company behind the app. We will help it respond.
Before a workspace is claimed, no customer stands behind it. We decide how its data is handled: we use it only to run and protect the service, and we delete it on the schedule in section 8.
We do not yet offer a data processing agreement. If you need one, write to [email protected].
4. What we collect, and when
We get personal data from you, from your devices and the software you run (including coding agents acting for you), from GitHub when you sign in with it, and from the apps that send traces to a workspace.
When you visit our website
- Usage analytics. We use PostHog to count page views and page exits: the page, the referring page, your browser and device type, and the country derived from your IP address. Our site sends these events to PostHog’s US region through our own domain, and your IP address reaches PostHog with them. The analytics use no cookies, and we don’t record clicks, typing or sessions, or identify you.
- Fonts. Our pages load a font from Google Fonts, so your browser sends Google your IP address and browser details.
- The browser demo. The demo at /demo runs a model in your browser. Your browser downloads the model from Hugging Face, which sees your IP address.
- The builders form. If you apply through /builders, we receive your name, email address, project description, the SDK you plan to use, your GitHub username if you give it, and whether you starred our GitHub repository. We store your application in Notion and notify our team in Discord.
- Hosting. Vercel hosts the website and receives standard request data, such as your IP address and browser details.
When you use the SDK or CLI without an account
- Model registry. When the SDK or CLI looks up a model, it calls our registry. By default it sends a header naming the SDK binding, the SDK and core versions, your platform and the inference backends built in. Setting the environment variable XYBRID_TELEMETRY_OPTOUT to 1 stops this header. Our registry records each lookup in PostHog’s US region with the model requested. A lookup made without an API key is identified by your IP address; one made with a key is identified by a hash of the key.
- Model downloads. Model files come from Hugging Face or from our own storage, and the host serving the file sees your IP address.
When a coding agent creates a workspace for you
When a coding agent runs xybrid connect without an account, it creates an unclaimed workspace. We receive:
- the name the agent reports for itself (for example “claude-code”), the CLI version, your operating system, processor architecture and SDK binding;
- the workspace name, which by default is the name of your project folder;
- an installation identifier and your IP address, which we store only as keyed hashes, to enforce signup limits: a hash of the identifier, a hash of the address (for an IPv6 address, of its /64 network) and a hash of a wider network range around it;
- the traces and gateway calls the workspace then sends, described below, and when it first and last sent them.
We also create an account for the agent itself, with a generated placeholder email address. We don’t ask for your name or email address at this stage.
When you sign in or claim a workspace
- GitHub sign-in. From GitHub we receive your verified email address, name, avatar, GitHub username and GitHub account number. We use GitHub’s access token once and don’t keep it.
- Sessions. We record your sign-in sessions, including your browser’s user-agent string.
- Claims. When you claim a workspace, we record who claimed it and when, the addresses of the Terms of Service and Privacy Policy shown to you, and the time you accepted them.
- Invitations. If someone invites you to an organization, we receive your email address and send the invitation through Resend.
Traces and telemetry
Trace export is off until a developer turns it on, and sending traces to us also needs an API key. Once it is on:
- Content. By default the SDK’s core includes content in traces: prompts and system prompts, model outputs and reasoning, and the arguments and results of tool calls, each up to a size limit. A developer can choose a metadata-only setting instead, which leaves the content out.
- Metadata. Traces record the model, timings, token counts, errors and similar details of each operation.
- Device details. Traces describe the device that ran the model: operating system and version, processor architecture and brand, and memory. A developer can add a device identifier and other attributes of their own.
We store traces in our analytics database. Content too large for it goes to our object storage. Separately, we combine telemetry from every workspace into reliability statistics by device class and model. These statistics have no workspace, user or device identifier, and every workspace can see them.
Gateway calls
Our gateways forward an app’s model calls to inference providers:
- The managed gateway (api.xybrid.dev) sends calls to third-party inference providers with our own accounts, or to our own inference service. We record each call’s model, token counts, timing, result and workspace. We currently don’t store the prompts or responses of these calls.
- The bring-your-own-key gateway (gateway.xybrid.dev) passes requests to the provider with the developer’s own key. It removes the end user’s IP address before forwarding and records the same metadata. In production we currently don’t store the request or response bodies. Our separate staging environment, which we use for testing, does store them.
The provider receives the full request either way. If an app’s own traces include the content of its gateway calls, that content is stored with the traces.
Logs and error reports
Our servers and load balancer log requests, including the IP address, time, address requested and browser or client details. The console and API send error reports to Sentry, and our Cloudflare services send logs to Better Stack.
When you contact us
If you email us, we receive your email address and whatever you write.
5. Why we use data, and our legal bases
European data protection law requires a legal basis for each use. Ours are:
| Purpose | Legal basis |
|---|---|
| Providing the services, your account and your workspaces | Performing our contract with you |
| Creating unclaimed workspaces, enforcing signup and usage limits, and preventing abuse | Our legitimate interest in offering a free service that stays available and safe |
| Security, logs, error reports and fixing problems | Our legitimate interest in keeping the services secure and working |
| Website and registry analytics | Our legitimate interest in understanding how our site and SDK are used, so we can improve them |
| Reliability statistics across workspaces | Our legitimate interest in giving every workspace better advice about which models run well on which devices |
| Reviewing builders applications | Taking steps at your request, and our legitimate interest in running the program |
| Sending invitations and service messages | Performing our contract with you |
| Meeting legal obligations and defending legal claims | Legal obligation, and our legitimate interest in protecting our rights |
You need a GitHub account to sign in or claim a workspace. Without one we can’t create your account. We don’t currently rely on consent. If we ever do, you can withdraw it at any time, and the withdrawal won’t affect what we did before.
We don’t use the content of your traces or gateway calls to train AI models. We don’t currently send marketing email.
6. Who receives data
We use these providers to run the services. Each receives only what it needs for its task:
| Provider | What for | Where |
|---|---|---|
| Google Cloud | Hosting the API, sign-in service and our inference service; our database and its backups; server logs | United States |
| Cloudflare | Our network, DNS and rate limiting; trace intake, the registry and the bring-your-own-key gateway; message queues; object storage | Global network |
| Tinybird | The analytics database that holds traces and usage records | United Kingdom |
| Vercel | Hosting the website, the console and the documentation | Global network |
| GitHub | Sign-in | United States |
| Resend | Sending organization invitations | United States |
| Sentry | Error reports from the console and the API | Germany |
| Better Stack | Logs from our Cloudflare services | Germany |
| PostHog | Website analytics and registry lookups | United States |
| Notion and Discord | Storing builders applications and notifying our team | United States |
Others receive data when you or your app ask for it:
- Inference providers. The managed gateway sends model requests to third-party inference providers with our own accounts. A provider can pass a request on to another provider that serves the model. The bring-your-own-key gateway sends them to the provider you choose, under your own account with the provider. Each provider handles the request under its own terms and in its own locations, which can be outside your country, for example in the United States or China. Write to [email protected] to find out which providers we currently use.
- Google Fonts and Hugging Face receive your IP address when your browser or device downloads files from them, under their own policies.
We may also disclose data to professional advisers, to authorities when the law requires it, or to a buyer or successor if our business is sold or merged.
We don’t sell personal data, and we don’t share it for cross-context behavioral advertising.
7. International transfers
We are based in the United States, and our main systems are there. As section 6 shows, providers also process data in the United Kingdom, Germany and on global networks, and inference providers in their own locations.
When we transfer personal data out of the European Economic Area or the United Kingdom, we rely on the European Commission’s standard contractual clauses, and the UK addendum to them, in our providers’ data processing terms. You can ask us for a copy at [email protected].
8. How long we keep data
We keep data for as long as we need it for the purposes in section 5, then delete it. Where we can’t give a fixed period, the table gives the criteria we use. Periods described as “currently” are settings we can change; we will update this policy when they change.
| Data | How long |
|---|---|
| Account details from GitHub, sessions and organization memberships | While your account exists. Ask us to delete it, as section 9 explains. |
| Organizations, projects and API keys | Until an owner deletes the organization or project in the console, or you ask us to delete them |
| Traces and telemetry | Most trace and telemetry records, including usage records of the managed gateway, are deleted automatically 90 days after they are recorded. Two tables, which hold usage records of the bring-your-own-key gateway and the spans sent through it, currently have no automatic expiry. |
| Trace content held in object storage | Until the workspace’s data is deleted, or earlier where our storage rules remove it |
| Reliability statistics across workspaces | 365 days. They carry no workspace identifier. |
| Signup hashes of the installation identifier and IP address | About a day after the workspace is created, or earlier when it is claimed |
| Website analytics and registry lookups | As long as we need them to understand trends |
| Server logs and error reports | As long as we need them for security and fixing problems, under our providers’ retention settings |
| Builders applications | Until we have decided on them and no longer need them for the program, or you ask us to delete them |
| Emails with us | As long as needed to answer you and keep a record of our exchange |
Unclaimed workspaces. Each workspace gets two deadlines when it is created:
- The key deadline, currently 7 days after creation. After it, the workspace’s keys stop working.
- The claim deadline, currently 14 days after creation. It can’t be set later than 14 days. After it, the claim link stops working and we delete the workspace.
Deletion runs in stages. It removes the workspace’s account records, keys and stored content. Traces the workspace sent, and usage records of its managed gateway calls, stay in our analytics database until they expire, currently up to 90 days. Usage records of its bring-your-own-key gateway calls, and the spans sent through that gateway, have no automatic expiry yet.
Some records remain after deletion: the agent’s reported name, the CLI version, operating system, architecture and binding, the workspace’s dates, hashes of its one-time tokens, and our records of the deletion and of any admin actions. We keep these to measure signups and protect the service. We don’t currently delete them on a schedule.
Deleting an organization in the console removes its account records but not its traces. Ask us and we will delete those as well.
Backups. Deleted database records currently stay in our database backups for up to 14 days. Records our analytics database rejects are kept by Tinybird for about a month, and messages waiting in our queues for a few days.
9. Your rights and choices
Depending on where you live, you may have the right to:
- access your personal data and get a copy of it;
- correct it;
- delete it;
- receive it in a portable format, or have us send it to another provider;
- restrict or object to how we use it, including uses based on our legitimate interests;
- withdraw any consent you gave.
To use any of these rights, write to [email protected]. There is no self-serve tool for these requests yet. We handle each request by hand and aim to answer within 30 days. To protect your account, we may ask you to confirm your identity, for example by writing from the email address on your account. When you ask for a copy, we will send it in a common machine-readable format.
If your data is in an app built with Xybrid, contact the company behind that app first, as section 3 explains.
You can also complain to a data protection authority, such as the one where you live or work, or the Information Commissioner’s Office in the UK. We would appreciate the chance to address your concern first.
You also have choices in the product:
- don’t turn on trace export, or choose the metadata-only setting;
- set XYBRID_TELEMETRY_OPTOUT to 1 to stop the SDK’s registry header;
- let an unclaimed workspace expire.
10. US state privacy rights
Residents of California and some other US states have rights under their state’s privacy law. This section sums up our practices in those laws’ terms.
What we collect. In the past 12 months we have collected:
- identifiers, such as your name, email address, GitHub username, IP address and API key hashes;
- internet activity, such as page views, registry lookups and server logs;
- device details, such as operating system, architecture and memory;
- the content of traces and gateway calls that apps send;
- professional details you give in a builders application.
Sources. You, your devices and the software you run (including coding agents acting for you), GitHub when you sign in, and the apps that send us traces.
Purposes and disclosures. We use the data for the purposes in section 5, and disclose it for business purposes to the providers in section 6.
No sale or sharing. We don’t sell personal information, and we don’t share it for cross-context behavioral advertising. We don’t use sensitive personal information to infer characteristics about you.
Your rights. You can ask to know what we collect and how we use it, and to access, correct or delete it. We won’t treat you differently for using these rights. You can send a request through someone you authorize, with proof that you authorized them. We will verify requests as section 9 describes, and answer within 30 days, or within the time your state’s law allows (45 days in California).
Retention. Section 8 describes how long we keep each category.
11. Cookies and similar technologies
Our website sets no cookies, and its analytics use none.
The console and sign-in service use cookies that are needed to sign you in and keep the console working. They don’t track you on other sites, so we don’t ask for consent to them:
| Cookie | What it does | How long |
|---|---|---|
| xy_idp | Keeps you signed in to the sign-in service (auth.xybrid.dev) | 30 days, renewed while you use it |
| console_session | Keeps you signed in to the console | 24 hours |
| console_oidc | Holds your place while you sign in | 10 minutes |
| xybrid_claim | Holds the claim link while you sign in to claim a workspace | 30 minutes |
| xybrid_last_workspace | Remembers the last workspace you opened | 1 year |
| sidebar_state | Remembers whether the console’s sidebar is open | 7 days |
| xybrid_user, github_oauth_state, github_oauth_next, github_oauth_return_to | Used by an older GitHub sign-in path, on xybrid.dev | Up to 30 days |
| xybrid_demo_mode | A demonstration switch for our team | 7 days |
The console and the documentation also save your light or dark theme choice in your browser’s local storage.
Do Not Track and Global Privacy Control. We don’t respond to Do Not Track signals. We don’t track you across other sites or sell or share your data, so there is nothing for these signals to turn off. We don’t let other companies collect data about your activity across sites through our services for advertising. Google Fonts and Hugging Face receive your requests when your browser loads files from them, as section 6 says.
12. Automated checks at signup
To keep free workspaces available, we limit automatically how many can be created from one network and in total each day, and the CLI must solve a small computing puzzle before signup. These checks use the hashes described in section 4.
If a check refuses a signup, the agent can try again later, or you can sign in with GitHub instead. A refusal only means a free workspace is not created at that moment. It has no legal effect on you or any similarly significant effect.
13. Security
We take reasonable technical and organizational measures to protect personal data. For example:
- the services are served over encrypted HTTPS connections;
- we store API keys, claim links, sign-in tokens and session tokens as hashes, not in readable form;
- we store signup identifiers only as keyed hashes;
- we don’t keep your GitHub access token.
No method of transmission or storage is completely secure. Keep your API keys and claim links secret, and tell us if they leak. If a breach affects your personal data, we will notify you and the authorities as the law requires.
To report a security issue, write to [email protected].
14. Children
The services are for developers and businesses. You must be 18 or older to use them. They are not directed to children, and we don’t knowingly collect personal data from children. If you believe a child has given us personal data, write to [email protected] and we will delete it.
15. Changes to this policy
When we change this policy, we post the new version here with a new “last updated” date. If a change is material, we will tell you at least 30 days before it takes effect, by email where we have your address or by a notice in the console. A change the law requires may take effect sooner. We keep previous versions and will send you one on request.
16. Contact us
Xybrid Inc.
251 Little Falls Drive
Wilmington, DE 19808
United States
Privacy: [email protected]
Legal: [email protected]
Security: [email protected]
We have not appointed a data protection officer, or a representative in the European Union or the United Kingdom. You can write to us directly at the addresses above.