No model training
Howard does not use prompts, files, conversations, tool results, or outputs to train shared AI models.
Security
Howard is designed to help engineering teams work with sensitive requirements, calculations, and design decisions with clear data boundaries.
Data protection
Howard does not use prompts, files, conversations, tool results, or outputs to train shared AI models.
Project content is processed to provide Howard, secure the service, resolve technical problems, and meet legal obligations—not for advertising.
Access is invite-only. Passwordless codes expire, login attempts are rate-limited, and session cookies are Secure, HttpOnly, and SameSite.
Conversation reads, writes, and deletion are scoped to the signed-in account. One user cannot request another user’s thread.
Users can delete individual conversations. Account and data requests can be sent to hello@howard.cx.
Traffic is encrypted in transit. Service credentials are loaded at runtime and are not exposed to the browser or committed with the application.
Enterprise review
Howard is designed for teams whose requirements, calculations, and design decisions are the product. Access is controlled, work is isolated to the account that owns it, and every answer keeps the evidence behind it.
Security review
Your account email, the questions you ask, files you attach, the conversation itself, and the engineering results Howard looked up to answer them. The Privacy Policy describes each category and why it is held.
No. Howard does not use prompts, files, conversations, results, or outputs to train shared AI models, and its model provider does not train on commercial API content.
Access is invite-only. Sign-in uses a single-use code that expires, attempts are rate-limited, and session cookies are Secure, HttpOnly, and SameSite. Every conversation read, write, and delete is scoped to the signed-in account inside the query itself, so one account cannot reach another’s work.
Howard stores customer data in its service environment. Traffic is encrypted in transit with TLS. Answering a question sends the content needed to Anthropic’s business API; login mail goes to SendGrid, which receives your address and never project content.
You can delete any conversation, which removes it and its messages immediately. Account-level deletion and data requests are handled directly by hello@howard.cx.
Email hello@howard.cx with “Security” in the subject. Reports route directly to the people who operate the service, not a queue.
The published Terms and Privacy Policy govern use of the service. Enterprise agreements, data processing terms, and a security addendum are handled directly as part of your review — bring your requirements and we will work to them.
We will walk through Howard’s architecture, data flows, current controls, and the requirements for your deployment.