Beta waitlist information.
When you request beta access, Shifu stores your email address, the optional reason you select for joining, and the time of your request. We use this information to manage beta access and send beta-access emails you agree to receive.
Demo requests.
When you request a demo, Shifu stores your name, email address, any optional phone number and notes you provide, a request identifier and the request time. We use these details to respond and arrange your demo. Submitting this form does not subscribe you to marketing emails. Please do not include confidential work content in the notes.
Resend processes the submitted details to deliver an internal notification to Shifu’s administrative inbox. We also store the notification’s delivery status so failed deliveries can be retried without losing your demo request.
We retain the request until you ask us to remove it. To request removal, email admin@useshifu.com from the address used for your demo request.
Coding session-review prompt.
When you request the session-review prompt, Shifu stores your email address, delivery consent, prompt version, request time and email delivery status. We use these records to deliver the requested copy and avoid duplicate sends. This request does not subscribe you to beta-access or marketing emails.
Resend receives the recipient address and prompt email for delivery. The review prompt runs in your existing coding tool and does not upload your raw coding session or generated report to Shifu. Your coding tool’s own processing and retention settings still apply. We retain the request and delivery record until you ask us to remove it; contact support from the address used for the request.
Website analytics.
The landing page uses Vercel Web Analytics to count visits and see aggregate traffic patterns without analytics cookies. We also use PostHog to connect signup-link clicks with login and activation in the app. PostHog stores a random browser identifier in a secure first-party cookie shared across our website and app, with local storage for analytics state. After you sign in, this identifier is linked to a pseudonymous account identifier so we can measure the journey through login, source connections and useful context. We do not send names, email addresses, form values or work content in these analytics events, and do not enable automatic interaction capture or session recordings.
You choose every connection.
Shifu shows the current access boundary for a source before you authorize it. Connect only sources you are allowed to use and that are relevant to the question you want to answer.
Work sync has clear boundaries.
Connecting a coding tool does not start capture. In the local setup wizard, choose projects, history and schedule settings, review the metadata preview, then select Start sync to authorize that scope. If you enable scheduled sync, the default interval is hourly; you can choose another interval or manual session approval. Preparation can read selected user messages and completed assistant reports and process them through your coding tool’s configured model provider. The original coding transcript is not uploaded to Shifu as a transcript file; derived records, evidence and metadata are retained. Redaction can miss sensitive information. Folder-file sync is off until you enable it and approve its folder in setup. Files you place directly in that approved folder upload as provided on scheduled sync and are redacted on the server. Folder-file processing runs separately from manual coding-session approval. See the full processing boundary.
GitHub access has a clear boundary.
During the current beta, the GitHub connection lets you connect your identity and choose repositories for future use. It does not read repository contents today.
Google data we access.
Google Meet is an optional connection, authorized separately from
signing into Shifu. When you connect it, Shifu requests
openid, email, and
https://www.googleapis.com/auth/meetings.space.readonly. We access:
- Your Google account identifier and email address, to identify the connected account and match your meeting contributions.
- OAuth access and refresh tokens, to maintain the connection and check for available transcripts.
- Meeting identifiers and dates, participant identifiers and display names, transcript identifiers and document links, and available transcript entries with text, speakers, and timestamps.
Shifu imports available transcripts for meetings that start after you connect. It checks periodically for delayed transcripts. Shifu does not join or record meetings, enable transcription, or download recordings or Google Drive documents through this connection.
How we use Google data.
We use meeting data to identify your contributions, extract work records with supporting excerpts, and provide your work history, summaries, and answers in Shifu. We match your Google identity to participant identifiers, label your turns, and redact other speaker labels and recognizable contact details before AI extraction. Other participants' words may be processed as context. Redaction reduces exposure but does not guarantee that every name, confidential detail, or personal identifier is removed. OAuth tokens are used for Google API access and are not included in AI prompts.
Google data sharing and AI processing.
Shifu's hosting and database service providers process Google data on our behalf to operate the service. Sanitized transcript sections are processed through Shifu’s configured, approved extraction model route. Relevant saved work records and excerpts may also be processed by the configured answer models and TypeSafe AI’s Jev to assess relevance and supporting evidence. The application integrates with OpenCode Go / Console and approved OpenRouter inference routes; TypeSafe judgments can run directly or through OpenCode. See providers and model routing for the application defaults and how to confirm the active deployment route. OpenCode's published Luna data policy states that inputs are not used for model training and that provider retention is 30 days. OpenCode's Jev policy states that inputs are not used for training. TypeSafe's Privacy Policy describes retention for as long as reasonably needed to provide its services or support its business purposes, without a fixed day limit. Its optional organization-level zero data retention (ZDR) deletes request content upon output and retains operational telemetry. TypeSafe must activate it for the applicable organization; do not assume it applies to Shifu’s active inference route without confirmation from admin@useshifu.com. These are hosted API integrations; processing is not limited to your device.
We do not sell Google user data or use it for advertising. We do not use or permit the use of Google user data, including raw, aggregated, anonymized, or derived data, to create, train, or improve foundational or generalized AI or machine learning models. Transfers are limited to providing the features you authorize, security purposes, legal obligations, or a business transfer with your explicit prior consent. Human access to Google data is limited to your explicit consent for specific data, security investigations, legal requirements, or aggregated and anonymized internal operations permitted by Google's policy. If you authorize an external assistant to retrieve your Shifu context, relevant retained records may be returned to that assistant.
How we protect Google data.
Shifu uses HTTPS for Google API and AI API transfers. Stored Google OAuth credentials are protected with authenticated AES-GCM encryption, using a key held separately from the database. OAuth authorization uses PKCE and single-use, browser-bound state. Authenticated, account-scoped access checks and database isolation protect meeting connections and imports. We limit the text submitted for AI processing and redact recognizable identifiers. These controls reduce risk; no system can guarantee absolute security.
Service hosting.
Railway hosts the Shifu API in US East (Virginia, USA). Contact admin@useshifu.com for the current full infrastructure provider and region inventory.
Google data retention and deletion.
- Google account connection details and encrypted OAuth credentials are retained while the connection exists. Disconnecting Google Meet deletes the local connection and credentials and stops future imports.
- Sanitized transcript sections are retained temporarily while processing or retries are pending. Their text is cleared when processing completes, reaches terminal failure, or you disconnect. Jobs with no progress for 30 days are eligible for hourly cleanup of abandoned processing text and retries, in bounded batches while the service is running successfully. A fresh upload can restart processing; extracted saved records are retained separately.
- Saved work records, review candidates, selected supporting excerpts, and source metadata remain in Shifu after disconnect so your work history remains available. They have no automatic expiry; you can request deletion using the contact below. Rejecting a record changes its review status and does not erase it.
- The AI provider's separate retention period for Luna requests is 30 days. Jev follows TypeSafe AI's retention policy, which has no fixed day limit under its standard terms. Optional ZDR applies only when confirmed for the applicable organization and route, as described above.
To request deletion of imported Google data, derived records, or your account data, email admin@useshifu.com from the email associated with your Shifu account and specify what you want deleted. We verify the request before deleting the relevant retained data. We will explain any legal retention requirement or remaining backup or service-provider retention that applies to your request. Disconnecting Shifu does not delete the original transcript held by Google. You can also remove Shifu's authorization from your Google Account connections.
Google API Limited Use.
Shifu's use and transfer of information received from Google APIs, including raw and derived Google Workspace user data, adheres to the Google API Services User Data Policy and the Google Workspace API User Data and Developer Policy, including their Limited Use requirements.
You can remove access.
Use Sources in the dashboard to disconnect a Google Meet or GitHub account or revoke a coding-harness key. This stops the current connection from being used. Disconnecting a source does not erase previously retained work records; contact us to request deletion.
Manual data deletion during the beta.
The Shifu team handles source-data and account-data deletion requests manually. Email admin@useshifu.com from the email associated with your account and tell us which source or account data you want removed. Do not include original work content or a connection key in the request.
We verify account ownership, confirm the scope, and coordinate stopping the relevant imports or sync before removing the applicable retained data and derived records. We confirm the result and explain any remaining legal, backup or service-provider retention, including the expected timing where available. Disconnecting a source stops its future use; deletion is handled through this separate request. Original source systems, local coding-tool history and external assistant conversations are managed separately. See data retention and deletion for details.
A submitted request or delivery receipt is not a deletion-completion confirmation. Submitting a request does not itself stop account access, imports or sync; we coordinate those steps while handling the request and confirm the result separately.
Generated work stays reviewable.
Shifu keeps the context behind summaries visible so you can check what is accurate and decide whether to use it. Source information remains in its original system, where any changes must be made.
Privacy questions and policy changes.
Contact the Shifu team at admin@useshifu.com for privacy questions, corrections, or deletion requests. We update this policy and its date when our practices change and notify affected users before materially changing how we use or share Google data.