User Flows
How users, developers, and agents interact with the .agt ecosystem — from claiming a name to programmatic resolution to agent-to-agent discovery.
Flow 1: Claim a Name
The registration site at /claim walks users through:
- User visits
/claim. - Types a name (e.g., "researcher"), clicks Search.
- API checks availability and shows the price.
- User enters a wallet address (
0x...). - Clicks "Claim this name."
- Backend creates the zone, triggers minting, and sets the
agt-version=1sentinel record. - Frontend polls minting status every 5 seconds.
- On success, the AgentConfigForm appears.
- User fills in: name, description, protocols, capabilities, and endpoints — or clicks "Skip for now."
- Backend writes the manifest. Today: legacy v0 TXT records. v1 will serialize signed JSON to IPFS and write a single
agt-manifest=ipfs://{cid}pointer (tracked in #110). - Done screen: domain claimed, config status, and next actions.
Flow 2: Browse an Agent from the Browser
- User types
exampleagent.agtin the browser URL bar. - TLD hint routes to the public .agt resolver.
- Resolver returns TXT records.
- Parser finds
agt-version=1and extracts allagt-*records into a manifest. - Frontend renders an agent card (not a webpage).
- User sees: name, description, protocols, capabilities, and endpoints.
- Clicking an endpoint opens it in a new webview tab.
Flow 3: Resolve Programmatically
Using the @agt/resolver SDK:
import { resolveAgent } from '@agt/resolver'
const agent = await resolveAgent('exampleagent.agt')
if (agent) {
// Use agent.endpoints[0].url to connect
// Check agent.protocols for supported communication methods
// Read agent.capabilities for what it can do
}Works in any JavaScript runtime. The SDK calls the public .agt resolver — no authentication or API keys required.
Flow 4: Agent-to-Agent Discovery
Future capability. The envisioned flow:
- An orchestrator agent needs a specific capability (e.g., web scraping).
- Queries the .agt registry for agents with capability
web-scraping. - Resolves manifests, compares pricing and reputation.
- Selects
scraper.agt, connects via MCP endpoint. - Delegates the task and receives the result.
Flow 5: Explore Agents
The agent directory at /explore:
- User visits
/explore. - Page fetches manifests from a seed domain list via
/api/agents. - Agent cards are displayed in a grid.
- User filters by protocol (MCP, A2A) or capability (research, chat).
- User searches by name or description.