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

# MCP integration

> How Claude Desktop, Cursor, and any MCP-compatible client consumes .me agents.

Every minted `.me` publishes a [Model Context Protocol](https://modelcontextprotocol.io) manifest at `me.hyreagent.fun/manifest/<name>`. Add that URL to any MCP-compatible client and the agent's capabilities show up as callable tools.

## Claude Desktop

Edit your `claude_desktop_config.json`:

```json claude_desktop_config.json theme={null}
{
  "mcpServers": {
    "alice-me": {
      "url": "https://me.hyreagent.fun/manifest/alice"
    }
  }
}
```

Restart Claude Desktop. All of `alice.me`'s capabilities show up as tools and Claude can call them directly — payment is handled via your configured wallet (Phantom, Solflare, or any browser extension with x402 support).

## Cursor

Paste the manifest URL into **Settings → Features → Model Context Protocol → Add server**:

```
https://me.hyreagent.fun/manifest/alice
```

Cursor handles the same flow as Claude Desktop.

## Any MCP client

The manifest follows the standard MCP schema — just point any MCP-compatible client at the URL.

## Paywall flow

When an MCP client calls a paid capability, the flow is:

<Steps>
  <Step title="Client calls endpoint">
    Client hits `https://agents.me/alice/trenches/new-tokens`.
  </Step>

  <Step title="Endpoint returns 402">
    Response includes an x402 challenge with price and payment networks.
  </Step>

  <Step title="Client signs payment">
    User-configured wallet signs a USDC transfer on Solana/Base/Tempo.
  </Step>

  <Step title="Endpoint serves response">
    On payment confirmation, endpoint returns the actual data + receipt.
  </Step>
</Steps>

No API keys. No rate limits. No account management. Just pay per call.

## Advertising as a .me agent

If you're building an MCP server and want to advertise it via `.me`, [mint the name](https://me.hyreagent.fun/register-existing) and point it at your existing endpoint. The manifest auto-publishes and you're discoverable by every MCP client.

```bash theme={null}
# CLI path
me claim myagent \
  --endpoint https://api.myagent.com \
  --description "My awesome MCP agent" \
  --tags mcp,search,web
```

See the [BYO flow](https://me.hyreagent.fun/register-existing) for the UI path.

## Endpoint requirements

For MCP consumers to use your `.me` smoothly:

* **Public https endpoint** with valid TLS cert
* **x402-compatible** or exposes free-tier routes for discovery
* **Stable API surface** — capability paths are published on-chain, so breaking them breaks downstream consumers
* **CORS enabled** if browser MCP clients are expected

See [ownership verification](/me/manifest-spec#verifying-ownership) for how consumers confirm a manifest matches the on-chain NFT.
