Embassy

• 

  Cover Page

• 

  page 2

• 

  workflow

• 

  Payments

• 

• 

  Calendar

Inspiration

Running a startup alone is exciting, but it can also be overwhelming. A solo founder has to handle product development, marketing, customer support, research, tools, subscriptions, purchases, and many small decisions often all at the same time. I realized that AI agents could help reduce this workload. An AI agent could research products, compare options, find the best tools, manage subscriptions, and even make purchases.

What it does

A user can simply describe a goal, such as: “Find the best project-management tool for my startup within my budget and set it up.” Embassy can understand the request, research available options, compare them, recommend the best choice, and take action on the user’s behalf. What makes Embassy different is its trust and policy layer. Before an AI agent performs a sensitive action or spends money, Embassy checks whether the action follows the user’s predefined rules, such as spending limits, approved merchants, allowed categories, and approval requirements.

How we built it

User Goal ↓ AI Agent Understands the Task ↓ Researches and Compares Options ↓ Embassy Trust & Policy Layer ↓ Prava Authorization ↓ Secure Execution ↓ Transparent Audit Log

Challenges we ran into

If the agent asks for approval before every action, it becomes slow and does not save much time. But if the agent receives unrestricted access, users may not feel comfortable allowing it to make important decisions. We addressed this by using policy-based autonomy. Users define their boundaries once, and the agent can work independently within those boundaries. If an action crosses a limit or violates a rule, Embassy either requests approval or blocks it.

Accomplishments that we're proud of

Building a working agentic workflow that can understand goals, make decisions, and take action Creating a trust layer that evaluates actions before execution Integrating secure mandate and passkey-based authorization through Prava Giving users control through customizable policies and spending limits Creating transparent logs for both successful and blocked actions Designing Embassy as a one-stop solution for solo founders who need support across multiple business tasks

What we learned

AI agents need clear boundaries when acting in the real world. User intent must be converted into enforceable policies. Automation should reduce work without removing user control. Trust improves when users can understand why an AI made a decision. A good AI agent should know when to act, when to ask, and when to stop. Security and user experience must be designed together rather than treated as separate features.

What's next for Embassy

We plan to expand Embassy so users can delegate a wider range of tasks, including: Researching and purchasing business tools Comparing SaaS subscriptions Managing recurring expenses Handling vendor purchases Booking travel and business services Automating operational decisions Coordinating multiple specialized AI agents

Built With

• ai
• css3
• html5
• linq
• prava
• pyth
• senso