AI projects

1. The customer asks their assistant for an appointment.
2. The assistant calls Vallon's MCP tools to find free times.
3. The customer picks a time and the booking is created in Vallon.
4. The center sees it on the iPad like any other booking.

Stack: MCP (TypeScript SDK), Vallon booking API, OAuth

1. A center owner asks a question.
2. The agent searches the product documentation and answers with the source.
3. For a technical issue, it asks which steps led to it.
4. It opens a ticket with what the team needs to reproduce the issue.

Stack: RAG with a vector database, Claude Agent SDK, evals with LLM-as-judge

1. A customer messages the center on WhatsApp.
2. The agent checks availability in Vallon.
3. It books the time and confirms in the same chat.
4. Anything it can't handle goes to the center's staff.

Stack: WhatsApp Cloud API, LLM tool use, Vallon API

Customer data is handled under Saudi Arabia's Personal Data Protection Law.

1. Each procedure is written as clear steps.
2. Code moves the customer from one stage to the next.
3. Inside each stage, an agent follows the procedure.
4. A team member approves the key steps.

Stack: State machine in code, LLM agents, human approval steps