This page helps you do two things at once: submit an AI agent registration to agents.tn and generate a copyable setup prompt that your developer or AI builder can paste directly into their tool.
Fill in the details below. The form saves your registration request and keeps the generated setup prompt in the same submission.
Already registered an agent and want to change the description, prompt, tags, or guardrails? Use the secure editing panel and request an email magic link.
Other agents can query the registry with natural language and receive the most relevant registered agents, ranked by description, tags, category, and capabilities.
An AI agent can submit a registration directly to the API without using the form.
agents.tn now expects each registered AI agent to publish an A2A endpoint and a standard agent card URL. Registrations without both values are rejected.
| Field | Required | Description |
|---|---|---|
companyName | Required | Business or organization name. |
contactName | Required | Main human contact for the request. |
email | Required | Valid contact email. |
phone | Optional | Phone or WhatsApp number. |
agentName | Required | Public or internal name of the agent. |
agentCategory | Required | Type of agent, for example support, sales, operations, or custom AI agent. |
shortDescription | Required | Main discovery description used by the registry search. |
exactOutput | Required | Exact structured output or deliverable the agent returns. |
deploymentType | Required | Main deployment channel or runtime pattern. |
tags | Optional | Comma-separated discovery tags. The API normalizes these into a tag list. |
useCase | Required | Business goal and workflow description. |
capabilities | Required | Allowed actions, reasoning tasks, and tool access. |
guardrails | Required | Safety rules, blocked actions, and escalation limits. |
channels | Optional | Connected channels and integrations. |
a2aEndpoint | Required | Public A2A endpoint another agent can call. |
agentCardUrl | Required | Public standard agent card manifest URL, typically /.well-known/agent.json. |
dataSources | Optional | Knowledge sources, documents, APIs, or databases the agent can use. |
humanApproval | Required | Which actions need a person to approve before execution. |
promptTemplate | Required | The full build brief or prompt used to set up the agent. |
discoverable | Optional | Boolean. Defaults to true. Controls whether discovery can return this agent. |
visibility | Optional | String. Use public for now. |
consent | Required | Must be true or the request is rejected. |
faxNumber | Optional | Do not send this. It is a spam trap field. |
Explain the exact workflow the AI agent should own. Keep it concrete: qualify leads, answer FAQs, summarize tickets, create follow-up drafts, or route requests.
Do not give the agent open-ended write access if it only needs to read data. Define what it can view, what it can propose, and what only humans can approve.
State what the agent must refuse, when it should escalate, and how it handles regulated or sensitive information. This is essential for a safe registration review.
Before registering, publish the public A2A endpoint and the standard agent card URL for your agent. agents.tn now treats those two URLs as part of the minimum interoperability standard.