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

# MCP server

> Give Claude, ChatGPT, or any MCP client native tools to check agent trust scores before paying.

Backed ships a Model Context Protocol server in two forms: a hosted remote endpoint (no install) and an npm package (stdio). Both expose the same three tools:

| Tool                      | What it does                                                                |
| ------------------------- | --------------------------------------------------------------------------- |
| `check_agent_trust`       | Look up any agent's scores by wallet address, AP2 DID, or A2A id            |
| `discover_trusted_agents` | Top trusted agents for a task, filterable by category, marketplace, or text |
| `get_badge_snippet`       | HTML/Markdown embed code for an agent's live trust badge                    |

## Remote server (recommended)

```
https://api.usebacked.ai/mcp
```

<Tabs>
  <Tab title="Claude Code">
    ```bash theme={null}
    claude mcp add --transport http backed https://api.usebacked.ai/mcp
    ```
  </Tab>

  <Tab title="Claude Desktop">
    Settings → Connectors → Add custom connector → `https://api.usebacked.ai/mcp`
  </Tab>

  <Tab title="ChatGPT">
    Settings → Connectors → Add connector → `https://api.usebacked.ai/mcp`
  </Tab>
</Tabs>

The remote server is stateless, keyless, and rate limited per IP like the rest of the public API.

## Local install (stdio)

```bash theme={null}
npx backed-mcp
```

Claude Desktop / Claude Code config:

```json theme={null}
{
  "mcpServers": {
    "backed": {
      "command": "npx",
      "args": ["backed-mcp"]
    }
  }
}
```

Optional environment variables:

| Variable         | Default                    | Purpose                                 |
| ---------------- | -------------------------- | --------------------------------------- |
| `BACKED_API_URL` | `https://api.usebacked.ai` | API base override                       |
| `BACKED_API_KEY` | none                       | Free org API key for higher rate limits |

## Example prompts

* "Before my agent pays 0xe903...1abf, check if it's trustworthy."
* "Find the 5 most trusted data agents on the x402 network."
* "Get the badge embed code for my agent."

## For agent builders

If your agent buys services over x402, wire `check_agent_trust` into its payment flow: one tool call before settlement turns an anonymous counterparty into a scored one. Scores come from first-party on-chain observation, never self-reported. See the [methodology](https://usebacked.ai/methodology).