Privacy Policy

Effective date: 2026-10-27

1. Information We Collect

  • Account data: Your email address and, when you use GitHub or Google sign-in, your provider user ID, display name, and the OAuth tokens obtained for that connection.
  • Account credentials: Provider API keys and Claude OAuth tokens you enter to run your agents. They are encrypted at rest and used only to run your account's jobs.
  • Repository data: Source code from repositories you connect, accessed only to perform the Service (propose fixes and open pull requests).
  • Job records: Title, description, status, error messages, and PR URL retained for each job you create.
  • Email address: Collected when you join the waitlist or create an account; used only for transactional and product notifications.
  • Agent registration data: A coding agent may submit an email address you provide, the project name, and your yes/no choice about limited setup or product follow-up. We immediately send a transactional claim email regardless of that choice.
  • Website inspections: The address of the website you ask Probie to inspect, the pages Probie visits, the screenshots and page errors it records as evidence, and the findings it writes. Inspection reports are public: each has its own URL and appears in Probie's report gallery and sitemap.
  • Requests from AI assistants: You can inspect public websites, install session capture, and operate your authorized Probie workspaces through an assistant such as ChatGPT using Probie's MCP server. Probie receives the inputs the assistant sends to its tools, including any descriptions, attachments, settings, or credentials you supply for the requested operation, plus client identification where supplied. It does not request your conversation history; text included in a tool input is received as part of that input.
  • Sign-in during an inspection: Only if you grant permission for that inspection, Probie may sign in to or create one account on the inspected website, using a Probie-controlled email address. Probie never asks for your passwords.

Connected apps

When you connect Claude, ChatGPT, or another MCP client, you choose the workspaces, projects, permissions, and access expiry. Connections can read and change authorized product data, including tickets, goals, settings, and integrations, subject to your current membership and the same workflow checks as the dashboard. Requested data and operation results are sent to the connected app. Credential setters do not return the stored secrets, but credentials supplied through an assistant pass through that assistant as tool inputs. We store authorization grants and hashed connector credentials to maintain the connection. Access tokens expire within one hour; refresh credentials rotate. Access expiry defaults to Never, or you can choose an expiry date. A Never connection remains available until you disconnect it or its authorization is otherwise revoked. Disconnecting stops future access; it does not remove information already returned to the connected app. We do not request your connected app conversation history or store connector tool arguments and results in application logs.

2. How We Use Your Information

We use collected data to: authenticate you; clone and analyse your repositories to propose code fixes; open pull requests on your behalf; inspect the websites you submit and publish the resulting reports; send the claim email for an inspection or agent registration; send transactional and product update emails; enforce usage limits; and improve the reliability and quality of the Service.

3. Data Processors

We share data with the following sub-processors to operate the Service:

  • GitHub: Repository access and pull-request creation. See github.com/site/privacy.
  • Anthropic: AI code generation when an Anthropic model is configured, and analysis of inspection screenshots and page content. See anthropic.com/privacy.
  • OpenAI: AI code generation when an OpenAI model is configured. See openai.com/policies/privacy-policy.
  • Amazon Web Services (SES): Transactional email delivery. See aws.amazon.com/privacy.
  • Akamai (Linode): Hosting, and storage of inspection screenshots and reports. See akamai.com/legal/privacy-statement.
  • Cloudflare: Network delivery and protection for the Service. See cloudflare.com/privacypolicy.

Inspection reports are public, so anyone with the link, or who finds them in Probie's report gallery or sitemap, can read them. AI assistants such as ChatGPT receive the tool results Probie returns to them.

4. Data Retention

Inspection reports stay public until you ask us to remove one at [email protected]. OAuth tokens are stored only while your account is active and are deleted when you close your account. Repository contents are processed transiently and are not persisted beyond the duration of an individual job. Job metadata (title, status, PR URL) is retained for debugging and audit purposes. Unclaimed agent-created projects, provisional telemetry, and registration records are deleted after seven days. We retain a non-reversible registration identifier to prevent an expired request from being reused. After claim, consent value, timestamp, and source are retained for audit; addresses without consent are used only for the claim transaction and support, while consenting addresses may receive limited manual setup follow-up.

5. Your Rights

You may request deletion of your account and associated data at any time by emailing [email protected]. We will respond within 30 days.

6. Security

We use administrative, technical, and organizational safeguards designed to protect your information. No system is completely secure; use the Service at your own risk.

7. Changes to This Policy

We may update this Privacy Policy from time to time. The effective date at the top of this page will reflect the latest revision.

8. Contact

Privacy questions or data-deletion requests? Email [email protected].