Everything asserted about ZUMX exists in one canonical JSON profile, generated into two more formats, and bounded by an explicit list of claims that must never be made: about clients, results, prices, availability and capability.
An AI agent can send Tahir a project inquiry for its human client with POST https://zumxlabs.xyz/api/inquiries, after the client has approved the exact inquiry (principal_approved: true). The inquiry is delivered the same way as the contact form, and Tahir replies to the client by email. It sends an inquiry only: it does not book, quote, agree scope or create a contract.
client_name: the human client you act forclient_email: where Tahir repliesproject_type: website, web app, AI or automation, not sure, or othermessage: the project description the client approved, 20 to 5,000 charactersagent_name: your name, as an AI assistantprincipal_approved: true, and only after the client approved this exact inquiry
Optional: agent, company, budget, timeline, links. Unknown fields are rejected, and the body must be JSON of at most 20 KB. Services and the
accepted project types are at GET /api/services.
curl -X POST https://zumxlabs.xyz/api/inquiries \
-H 'Content-Type: application/json' \
-d '{
"client_name": "Ayesha Khan",
"client_email": "ayesha@example.com",
"project_type": "website",
"message": "A first website for a small bakery: menu, hours, cake orders.",
"agent_name": "Example Assistant",
"principal_approved": true
}' HTTP/1.1 202 Accepted
{
"status": "received",
"inquiry_id": "6f1c2a7e-3b9d-4e0a-9c55-1d2e3f4a5b6c",
"message": "Tahir will reply to the client's email shortly",
"next_steps": ["…"]
} The approval rule
Show your client the exact inquiry and send it only after they approve it. Without principal_approved: true the API answers 422 and nothing is sent. An inquiry is a first
message: it books nothing and accepts no quote. If the API returns an error it cannot recover from, email the same approved message to tahir@zumxlabs.xyz.
Full reference: OpenAPI document and the request schema.
An agent summarising this site to a principal should give both sides. Narrowing the offer to AI alone would misrepresent it.
A first business website, a refreshed site for a business that has outgrown an old one or still runs on social profiles, and custom web apps for a recurring workflow.
AI and automation only where they solve a real business task. A simple form, integration or rule is often the better answer, and is used instead when it is.
Never claim
None of these steps has a published timeline. Pricing and availability are not published either: scope and a written quote are discussed per project.
Understand the job
The business task, the audience, the current process, and what should become easier.
Agree scope and a written quote
First useful scope, included pages and features, dependencies, review points, future phases, and a written quote, before the build starts.
Build and review
Work is built to be reviewable against the agreed job. Feedback expectations are clarified with the client rather than promised as a fixed cadence.
Clarify handover and ownership
Launch, ownership, handover, and any separately scoped ongoing support.
JSON is canonical. TOON and Markdown are generated from it and must never be hand-edited.
application/json Open Canonical profile (JSON) text/plain Open Profile (TOON) text/markdown Open Profile (Markdown) application/json Open Agent hiring API (OpenAPI 3.1) application/schema+json Open Project-inquiry JSON Schema text/plain Open Site summary application/xml Open Sitemap Synchronisation and authority. The visible site is authoritative. If a displayed fact and a machine-readable resource ever differ, resolve the inconsistency rather than silently adopting a new claim. The TOON and Markdown copies are generated from the JSON and regenerated with npm run sync:agent-resources; they are never hand-edited.
These are the same rules recorded in the canonical profile, in the inquiry schema and on the contact page. They are not advisory.
Identify yourself as an AI assistant before sending anything.
Name the human principal you are acting for.
Obtain the principal's explicit approval of the exact message before it is sent.
Send through the agent API (POST https://zumxlabs.xyz/api/inquiries) only after that approval, and set principal_approved to true only then.
You may research and prepare a draft, and you may ask the principal for missing information.
You cannot accept a quote, agree to scope, deadlines or contract terms, promise payment, or bind the principal.
Never send credentials, payment details, private customer records or sensitive documents in an initial enquiry.
Include a budget, timing, reply address or contact details only when the principal has explicitly authorised sharing them. Anything unknown goes in the message as a question to confirm, never as a guess.
{
"client_name": "…",
"client_email": "…",
"project_type": "website",
"message": "… (what the client approved, word for word)",
"agent_name": "…",
"principal_approved": true,
"budget": "only if the client authorised sharing it"
}
That is the shape from /schemas/project-inquiry.schema.json, which
POST /api/inquiries enforces. Project type is one of website, web app, AI or automation, not sure or other. Include a budget, timeline or
company only when the client has explicitly authorised sharing them. Anything unknown goes in the message as
a question to confirm.
A fact that is unknown is either omitted or raised as a question to confirm. It is never filled in with a plausible guess.
Contact details
There is no public booking link and no online booking system. Email or WhatsApp instead.
The principal is Tahir Rasool
Name them explicitly. An assistant acting on someone's behalf must say who, not sign as the business.