---
title: "Start here"
description: "Roles, vocabulary and limits before you connect an agent or accept agent signups."
image: "https://docs.agentregistrar.dev/og/docs-default.png"
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.agentregistrar.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Start here

Agent Registrar connects a verified human owner, an agent sub-identity and a relying party that accepts brokered signups. This section explains the roles, the handshake and the limits before you connect a harness or implement a provider.

## Pick your job

- **You run an agent.** Connect a harness to the broker, register the agent, complete a first brokered signup and keep control of the account. Start with [Quickstart - use an agent](/start/quickstart-agent/).
- **You run a service.** Publish an identity-only discovery document, verify a signed assertion and return a receipt. Start with [Quickstart - accept agent signups](/start/quickstart-provider/).
- **You want to try the deployed test network.** Use the [demo network](/demo/) to compare same-broker, cross-broker and unverified signups.
- **You need the contract.** Read [Protocol](/protocol/) and the [specification v1.2.0](https://github.com/agentregistrar/spec/blob/v1.2.0/agent-identity-v1.md).

## Read first

- [What is Agent Registrar](/start/what-is-agent-registrar/): the product, the protocol and the broker, and what each one owns.
- [Key concepts](/start/concepts/): human owner, agent, harness, broker, relying party, assertion, receipt, discovery, context guide, ledger, kill switch.
- [How it works](/start/how-it-works/): the end-to-end handshake with a diagram and the claims a provider sees.
- [Trust and limits](/start/trust-and-limits/): the liability chain, what revocation does, and the known gaps.

## What this site is not

The docs explain behavior and point at the canonical source for every rule. They do not restate normative text, and they are not a substitute for the specification when you implement a relying party.

> **Where the rules live**
>
> The protocol rules are in the [specification](https://github.com/agentregistrar/spec/blob/v1.2.0/agent-identity-v1.md). Broker and MCP behavior is documented from the deployed service and the shared tool schemas; each behavior page names what it was checked against.

Source: https://docs.agentregistrar.dev/start/index.mdx