Privacy Policy
Last updated: 10 Nov 2026
This policy explains how personal data is handled when you use the PhillipCLI hosted service: the Tara Portal, the Tara registry and the Ryoma metrics service, together with the Phillip command-line tool and VS Code extension when you connect them to the hosted service. For the technical detail of telemetry, see also Data collected.
1. Who is responsible
The controller of the personal data described here is <legal name of the seller: full legal name if individual, registered company name if company>, Vietnam. Contact for privacy questions and requests: <contact email for legal and privacy requests>.
For the content an organization puts into the Service (resources, Knowledge base documents, project data), the organization decides what to upload and who may see it. We process that content for the organization to provide the Service.
2. What we collect
Account data. When you sign in, Tara stores your handle (your Tara login), name, email address, avatar URL, whether your email is verified, the time of your last sign-in, and the link between your account and the sign-in provider (provider name, your user ID and login at that provider). Tara asks GitHub only for your basic profile and email (read:user, user:email) and GitLab only for openid, profile and email. It never asks for access to your repositories when you sign in. With GitLab or your organization's single sign-on, Tara also reads your group memberships, only to place you in the right organization. With single sign-on, Tara reads the login, email, name and group claims your identity provider sends. If you sign in with Google, Tara receives your basic profile and whether your email is verified.
Email sign-in links. If you use an email link, Tara stores your email address, a hash of the link, the IP address of the request (to limit abuse), and the times of creation, expiry and use. The link expires in 15 minutes and works once. These rows are deleted when a new link is created more than one day after they were made. We send the link through an SMTP email service.
Sessions and tokens. Tara stores only hashes of session and API tokens, never the tokens themselves. Browser sessions last 7 days and Phillip sessions 12 hours by default. API tokens record a name, scopes, creation, expiry and last-used time.
Organizations. Organization name and slug, members and roles, invitations (email), verified email domains, projects and project members, and, if you set it up, single sign-on settings (the client secret is stored encrypted). Tara also keeps an audit log of administrative changes: who did what to what, and when.
GitHub and GitLab connections. If an organization installs the GitHub App or connects GitLab, Tara stores the installation or connection, the list of repositories and organization members it needs to place people and import what you choose, and webhook delivery identifiers. It does not copy source code except for repositories and paths an organization explicitly imports as resources or Knowledge base sources.
Resources and Knowledge base. Resource packages, versions and files, signatures, check results, download counts per version per day, and Knowledge base documents and their text chunks, as uploaded by the organization or person. Public resources are visible to every user, together with the owner's name.
Billing. Paddle takes your payment details. Tara stores only the identifiers and records that Paddle sends back: the Paddle customer ID, subscription ID and price, the billing email, the plan, quantity and status, and for each payment its amount, currency, status, dates, invoice number and period, plus the webhook events received from Paddle. Tara does not receive or store card numbers.
Usage telemetry (Ryoma). When you run Claude Code through ph and are signed in to Tara, Phillip sends OpenTelemetry data to Ryoma. Ryoma's collector keeps only an allowlist of fields: session ID, request ID, model, token counts (input, output, cache), estimated cost, the names of skills, plugins and agents used, the repository address, the task ID taken from the branch name, Guardrails decisions (rule ID, decision, severity, hook, note), and the user and organization identity that the server attaches from your token. Prompts, responses, tool inputs and outputs, shell commands, error text, your Claude email and account ID, and any other field are dropped before anything is written. Phillip does not turn on the Claude Code settings that record prompt or response content. Tool parameters may travel to the collector but are removed there before storage. Ryoma also reads task data (titles, labels, owners, timestamps) from GitHub or GitLab to compute Flow metrics.
Notifications. If a notification channel is configured (for example Telegram, Slack, Zalo or email), messages carry task metadata such as the task ID, title and link.
Support. If you write to us, we keep your message and our reply.
3. What stays on your machine
Phillip runs on your computer and starts Claude Code there. Your source code, prompts and files are read by Claude Code locally and are not sent to PhillipCLI by Phillip. Claude Code itself talks to Anthropic using the Claude account or API key you signed in with; that is your own relationship with Anthropic and is governed by Anthropic's terms and privacy policy. Phillip sends to us only: your sign-in, the packages you download or publish, Knowledge base operations, and the telemetry in section 2. Bandum, the acceptance tool, runs on your machine with SQLite or PostgreSQL in Docker. Its data stays there, and our services do not call into your machine. If you publish a resource or upload a document to Tara, that content does leave your machine, as you intended.
4. Why we use it
- To provide the Service: sign you in, apply roles and policies, deliver resources, show metrics (performance of the contract).
- To bill: confirm payments, keep seats right, prevent duplicate charges, issue refunds (performance of the contract, legal obligations).
- To keep it safe: rate-limit sign-in, scan resources for malware, keep an audit log, investigate abuse (legitimate interests).
- To answer you and improve the product (legitimate interests).
We do not sell personal data and we do not use it for advertising.
5. Who processes it
We use these service providers to run the Service:
| Provider | Why | What it receives |
|---|---|---|
| Paddle.com (Merchant of Record) | Checkout, payment, tax, invoices, refunds, billing support | Your billing details, which you give to Paddle directly; Paddle is an independent controller for payment and tax data |
| GitHub | Sign-in, GitHub App, importing resources and tasks | What you authorize at GitHub |
| GitLab | Sign-in, GitLab connection | What you authorize at GitLab |
| Sign-in, if you choose it | Your sign-in with Google | |
| Anthropic (Claude API) | Automatic review of resources before they are made public, and of new versions of public resources | The resource name, and the file paths and text content of the resource package and its attached documents; nothing else from your account |
| <transactional email provider> | Sending sign-in links and invitations by email | Your email address and the message |
| <embedding provider for Knowledge base semantic search, or "none (not enabled)"> | Semantic search in the Knowledge base, only where enabled | Text of Knowledge base documents |
| <hosting provider> | Hosting the Service and its databases | All data described in section 2 |
Your organization may also connect its own tools, such as chat channels for notifications or its own identity provider. Those are chosen and controlled by the organization.
We may disclose data if the law requires it. If our business is transferred, the data may go with it, and we will say so.
6. How long we keep it
Retention periods set in the Service: email sign-in link records are deleted when a new link is created more than one day after the record; sign-in codes for the CLI and OAuth are deleted one hour after they expire; billing idempotency keys are deleted after 24 hours; browser sessions expire after 7 days and Phillip sessions after 12 hours by default; invitations expire after 7 days.
For everything else, including account data, organization data, resources, audit logs, billing records and telemetry, we keep it while your account or organization exists, and we delete it when you ask (see section 7). We may keep what the law requires us to keep, such as invoices, and Paddle keeps its own records under its policy. Telemetry in Ryoma has no fixed retention unless the operator sets one.
7. Your rights
You can ask us for a copy of your personal data, to correct it, to delete it, or to export it. Write to <contact email for legal and privacy requests> from the email address on your account. We will answer within a reasonable time and may need to confirm who you are. You can change your name and email at your sign-in provider, and your organization's Owner can remove you from the organization. If you are in the European Economic Area or the United Kingdom you also have the rights to object to or limit processing and to complain to your data protection authority. Deleting an account may remove access to organizations and resources you own, so transfer ownership first. Your organization's Owner decides about the organization's own content.
8. Cookies and similar technologies
This documentation site sets no cookies, has no analytics, and loads nothing from outside sources. The Tara Portal uses only essential cookies: a session cookie (tara_session) to keep you signed in, and short-lived cookies during sign-in (tara_oauth_state, tara_oauth_next, tara_oauth_verifier, 10 minutes). They are HttpOnly, and Secure on HTTPS. The Portal keeps your language choice in your browser's local storage. When you open Paddle's checkout, Paddle runs its own window and may set its own cookies, under Paddle's policy.
9. International transfers
The Service and the providers above may process data in countries other than yours. Where the law requires it, we rely on safeguards such as standard contractual clauses.
10. Security
Tokens and codes are stored as hashes, sensitive secrets such as single sign-on client secrets are encrypted, data of each organization is isolated at the application and database level, packages are signed, and the connection to the Service uses HTTPS. No system is perfectly secure, and we cannot promise otherwise.
11. Children
The Service is for professional use and is not meant for children. We do not knowingly collect their data.
12. Changes
We will update this page and its date when this policy changes, and give reasonable notice of material changes, for example by email or a notice in the Portal.