Effective date: October 20, 2026
VibeKit ("we", "us", "our") operates the VibeKit platform, the website at vibekit.bot, the iOS application, the web dashboard at app.vibekit.bot, the CLI, and published npm packages (collectively, the "Service"). VibeKit is the data controller for the personal data described in this policy; contact details are in Section 11. This policy explains what data we collect, how we use it, who we share it with, and your rights.
By using the Service you also agree to our Terms of Service.
sk-ant-api03-, Anthropic OAuth tokens sk-ant-oat01-, OpenAI API keys sk-proj-, or ChatGPT-subscription Codex OAuth tokens), they are stored encrypted at rest and used solely to call the relevant provider on your behalf. See Section 3 for details.<name>.vibekit.bot.Legal bases (EEA/UK). We process account, workspace, transcript, payment, and usage data to perform our contract with you (operating the Service and billing it). We process server logs, security events, the automated quality review, abuse investigation, and product-improvement analysis under our legitimate interest in running a secure, working service, balanced against your rights. Analytics cookies in the EEA, UK, and Switzerland, and non-essential email, rest on your consent, which you can withdraw as described in Sections 2 and 6. Where law requires it, we retain billing records to meet legal obligations.
Data is stored on AWS infrastructure in us-east-2 (Ohio). All connections to the Service use HTTPS/TLS.
BYOK credentials and environment variables are encrypted at rest with AES-256-GCM. Each user's data is encrypted with a per-user key derived from a master key via HKDF-SHA256 with the account UUID as salt, master-key compromise alone does not expose any single user's data. Plaintext credentials are only held in process memory at request time and are never written to logs, telemetry, analytics, or backups. We redact known secret formats (sk-, sk-ant-, sk-proj-, GitHub tokens, JWTs, and others) from any text that gets logged.
Each app runs in its own AWS Fargate task with an isolated workspace. Per-tenant shell access is sandboxed by a bwrap wrapper at the OS layer. Database access is gated by Postgres row-level security so users only read and write their own rows.
No system is perfectly secure. If a breach affects your personal data, we will notify you and the relevant authorities as required by applicable law.
Your apps' own visitors. Apps you build and host on VibeKit may collect data from their own end users (signup forms, databases, email features). That data belongs to you, the app owner, and is processed by VibeKit only as your hosting infrastructure. You are responsible for your app's own privacy disclosures and legal compliance toward its users; this policy covers VibeKit's users, not the visitors of apps our users operate.
VibeKit routes your AI requests to third-party providers. The routing depends on whether you've configured BYOK:
Media-generation exception: Music and video generation are currently platform-credit services and use OpenRouter plus the serving provider even if you have a BYOK credential for chat. Video generation is asynchronous and is not eligible for OpenRouter Zero Data Retention; the provider temporarily retains the generated output so it can be retrieved.
Instant tools: When on for an agent (the default for accounts with an eligible paid credit purchase), these optional tools use Composio's provider accounts, separately from model BYOK. Tool inputs pass through VibeKit and Composio to the selected upstream service. Results returned to the agent can then enter the context sent to its active AI provider under the routing above. See Section 5a for consent and retention.
We do not send your email address, payment information, VibeKit account password, or BYOK model-provider credentials to AI providers. The exception is a credential intentionally returned by a connected-account tool after the separate confirmation described in Section 5.
Anthropic and OpenAI do not train models on API data, on either the BYOK or platform-credit route. OpenRouter, the aggregator we use for the non-BYOK route, may use prompts and responses to improve their product under the data-discount setting we've enabled (see Section 4a above); they have stated this is not used for model training. Free-model pool operators are an exception: the third parties serving free-tier requests set their own data practices, and some may use inputs to improve or train their models, if that matters for your content, use a paid provider or BYOK. Data sent to AI providers is never used for advertising by us.
Before first use of the AI agent, you're asked to acknowledge data sharing with third-party AI services. You can withdraw consent by removing your BYOK credentials and discontinuing use of the agent. Instant tools require a separate owner acknowledgment before enablement; setting the agent's Instant limit to zero stops new Instant calls.
If you use Connections to link a third-party account (for example Gmail, Slack, Notion, GitHub, Jira, HubSpot, Stripe) to one of your workspaces, the OAuth flow is operated by our integration provider, Composio. Composio receives and stores the access token for that account; VibeKit does not receive your account password or the OAuth token used to establish the connection. What we store is a bookkeeping record: which workspace is linked to which application, the connection's status, and timestamps.
Data that flows through a connected account. When your agent uses a connection, the request is made server-side by VibeKit through Composio to that provider, and the provider's response is returned to your agent so it can act on it. That content, for example the body of an email, the rows of a spreadsheet, or the text of an issue, therefore passes through Composio and VibeKit, and then into your agent's context, which means it is also sent to the AI provider handling that request under Section 4 above. Some provider tools can intentionally return sensitive credentials such as API keys, access tokens, passwords, or database connection strings. Before running one, the Service requires a separate, explicit confirmation phrase from you and warns that the result will enter the agent transcript and the active AI provider's context. VibeKit records that a sensitive tool was confirmed and executed, but does not separately log its arguments or returned value. Do not connect an account, or approve a sensitive credential request, unless you are willing to have that content processed by an AI model. We do not separately retain the content of connected-account responses beyond the agent transcript for that conversation, which follows the transcript and archive rules in Sections 1 and 7. Composio, however, keeps its own log of each request your agent makes through a connection, including the request and the provider's response (and so any sensitive credential you approved), for up to one year under its data-retention policy. Those logs are identified by an internal VibeKit user ID, not your name or email, and deleting the workspace or your VibeKit account does not remove them.
Actions. Connections can write, not just read. Actions your agent takes, such as sending an email or creating a record, happen inside the provider's systems and are recorded by that provider under its own policy. We log the workspace, the application, and the name of the action taken, so the activity is auditable; we do not log the contents of what was sent.
You can disconnect any account at any time from that workspace's Connections screen. Disconnecting revokes the agent's access going forward and asks Composio to revoke the stored token. Data the provider already holds, and actions already taken, are governed by that provider's policy. Composio's own privacy policy is available at composio.dev/privacy.
Instant tools run on Composio's provider accounts, so you do not need to connect a personal account or supply that service's credentials. Inputs can still include personal information: for example, a search query, a URL, a document, or details you ask a tool to process. The selected tool's inputs pass through VibeKit and Composio to its upstream provider. Results returned to the agent can be included in the conversation and sent to the active AI provider. Instant is on by default for agents of accounts with an eligible paid credit purchase, and the account owner can turn it off for any agent in that agent's Settings. Do not submit information you are not authorized to share with those services.
VibeKit storage. To recover an interrupted response without paying for another call, we temporarily store received tool results that fit the recovery size limit for retrieval for 24 hours from the original call, then remove them through our periodic cleanup. Deleting the workspace or account removes this recovery copy. This is separate from any result already included in an agent transcript or saved in your workspace, which follows the retention rules in Sections 1 and 7. We retain the call identifier, tool, status, timing, price and billing evidence needed to reconcile charges and handle disputes; the financial record does not retain the tool's full inputs or output. Those financial records can remain after workspace or account deletion.
Composio and upstream storage. Composio's published Instant guidance says it does not store Instant request or response payloads. Execution metadata remains subject to its retention policy, which keeps execution logs for up to one year. Files used or returned by a tool may be staged by Composio for 24 hours, and download links may expire sooner. The upstream provider can retain data under its own policy; Instant is not a promise of end-to-end zero retention or no model training. Deleting a VibeKit workspace or account does not delete records already held by Composio or an upstream provider. Review the selected service's policy before submitting sensitive information.
Turning it off. Turn Instant off in that agent's Settings, or set its Instant spending limit to zero, to stop new Instant calls. This does not cancel calls already dispatched or remove data already sent. Disconnecting a personal account from Connections does not disable Instant, because Instant uses separate provider accounts.
We use Google Analytics (GA4) to measure aggregate traffic, both on the marketing pages at vibekit.bot and on the signed-in web dashboard at app.vibekit.bot. On the dashboard this measures product usage (which pages are visited, and aggregate patterns) so we can see where the product is confusing or broken; we do not send your prompts, file contents, agent transcripts, or credentials to Google Analytics. The iOS app contains no third-party analytics at all. We use a small number of first-party cookies for session management and one (vk_referral) to attribute referral signups.
Consent. If you are in the EEA, the UK, or Switzerland, analytics and advertising storage are set to denied before anything loads, and stay denied unless you accept in the banner we show you. Declining is a real choice that we remember: Google Analytics then sets no cookies and records no advertising identifiers for you, and the rest of the Service works exactly the same. You can change your mind by clearing this site's data in your browser, which brings the banner back. Outside those regions analytics runs by default, and the opt-outs below are always available to everyone.
Google signals (Advertising Features). Our GA4 property has Google signals enabled, and because Analytics runs on the dashboard as well as the marketing pages, this applies to signed-in visits too. For visitors who are signed in to a Google account and have consented to ads personalization, Google may associate their visit with Google's own account information, which can include location, search history, YouTube history, and data from sites that partner with Google, to give us cross-device and aggregate demographic/interest reporting. We only ever see this in aggregate (e.g. age brackets, gender, interest categories); we never receive data that identifies an individual, and we do not use it for ad targeting. We adhere to Google's Advertising Features policy. You can review or turn off this association at any time via Google Ads Settings and My Activity, or opt out of Google Analytics entirely with the Google Analytics opt-out browser add-on.
Agent transcripts and workspace data are retained while your account is active. You can delete a workspace (its files, env vars, uploads, and any Fargate task) from the dashboard at any time; the archived chat described in Section 1 is kept, as described there. Server logs are kept for 10 days. Audit-log entries (security events) are retained for 12 months. Hosted-app logs shown in the product are kept for 1 hour on Free, 24 hours on Builder, and 7 days on Pro.
Account deletion removes all associated data within 30 days, with three exceptions: billing and transaction records needed for tax, accounting, reconciliation, and disputes, including the minimal Instant call records described in Section 5a; your archived chats and their quality-review rows (Sections 1 and 2), which are kept without your user ID or workspace address so they are no longer linked to your account; and a one-way hash of the sign-in identifiers on the deleted account (email, and the Apple, Google, GitHub or Telegram ID if present). The hash cannot be turned back into the identifier and is used only to stop the same person collecting the free signup credit again by deleting and re-creating an account. Deleting your account also cancels an active Stripe subscription immediately; an active Apple subscription must be cancelled with Apple first. Composio's connected-account logs and Instant execution metadata (Sections 5 and 5a) follow its own retention policy and are not removed by deleting an app or your account. The temporary Instant recovery copy follows the separate 24-hour retrieval and deletion rules in Section 5a.
Depending on where you live, you may have rights under the GDPR (EEA/UK), CCPA/CPRA (California), or similar laws. You can:
We do not sell or share your personal information as those terms are defined under the CCPA/CPRA, and we do not discriminate against you for exercising any privacy right.
VibeKit infrastructure is in the United States (AWS us-east-2). If you access the Service from outside the US, your data will be transferred to and processed in the US. We rely on AWS's GDPR-compliant data-processing terms and Standard Contractual Clauses where applicable.
The Service is not intended for users under 13. We do not knowingly collect data from children under 13. If you believe we have collected data from a child, contact [email protected] and we will delete it.
We may update this policy. Material changes will be communicated through the platform with at least 30 days' notice for paying users where reasonable.
Privacy questions and general support: [email protected].