What am I building here?
You are building an agent that helps a person with a specific task. Hussh helps that agent ask before using personal data, receive only what the user approves, and keep a record of what happened.
FAQs
Plain-English answers for building with Hussh.
Use this page when you are trying to understand what to build, why permission matters, and how to get from an idea to a tested agent.
What Hussh Is
Is Hussh a model provider?
No. You can use your own model, runtime, cloud, local setup, or app surface. Hussh provides the consent, data-access, testing, and review layer around the agent.
What is Agent ONE?
Agent ONE is the user-facing assistant. It can call your agent when the user needs the job your agent is built for.
What is Nav?
Nav is the permission layer. It asks the user before personal data is shared and makes sure the agent only receives the approved data.
Building An Agent
Where should I start?
Start in Builder. Describe what your agent should help with, choose where it runs, and say what personal data it needs.
Do I need to understand the whole protocol first?
No. The SDK and Builder are meant to handle the consent flow for you. You should understand what data your agent needs, why it needs it, and what it should do if the user says no.
Can my agent run outside Hussh?
Yes. Your agent logic can run in your own environment. Hussh controls the request, permission, approved data, revocation, and audit trail around that work.
Can I use MCP from my coding tool?
Yes. Hussh MCP is designed so tools like Codex, Cursor, Claude Code, Windsurf, and other MCP-capable tools can scaffold, validate, test, and package an agent.
Data And Permission
Does my agent get all of a user’s data?
No. Your agent only receives the fields the user approves for the stated reason. If the user approves less, your agent must continue safely with less information.
What happens if a user says no?
Your agent should still be useful. It can explain what it can do without that data, ask for a narrower set of data, or continue with a safer generic answer.
What happens if a user revokes access?
The agent must stop using that data. Sandbox tests should prove that revocation is handled cleanly before the agent is submitted for review.
Does Hussh store the user’s personal data?
The product goal is that Hussh routes and protects access rather than becoming a warehouse for personal data. Agents receive approved context for the approved purpose.
Testing And Review
Why do I need the sandbox?
The sandbox lets you test with fake users and fake personal data. It shows whether your agent asks clearly, handles denied access, stops after revocation, and produces useful output.
What does “ready to submit” mean?
It means the agent has a clear purpose, asks for only the data it needs, passes validation, works in sandbox runs, and has evidence reviewers can inspect.
Can I publish immediately?
Not yet. A developer-built agent needs to pass the sandbox and review path before it can receive real user context through Hussh.
What should I do after my first agent works?
Run more personas, inspect the records, tighten the data request, download or package the agent, then prepare it for review.