# TrustMesh MCP Server

Trust scoring for AI agents via Model Context Protocol.

[![npm version](https://img.shields.io/npm/v/trustmesh-mcp.svg)](https://www.npmjs.com/package/trustmesh-mcp)

## What is this?

An MCP server that gives AI agents (Claude, Cursor, Windsurf, etc.) the ability to **check trust scores** and **rate interactions** with other agents — before transacting.

## Why?

When autonomous agents trade APIs, data, and services, they need to know: *can I trust this counterparty?* TrustMesh provides the reputation layer that's missing from the AI agent ecosystem.

## Tools

| Tool | Description |
|------|-------------|
| `trustmesh_rate` | Rate an interaction with another agent |
| `trustmesh_score` | Check trust score before transacting |
| `trustmesh_batch_score` | Batch check multiple agents |
| `trustmesh_dispute` | File a dispute against a bad rating |
| `trustmesh_resolve_dispute` | Resolve a dispute |
| `trustmesh_graph` | See who trusts whom |
| `trustmesh_peers` | List all known peers |

## Install

```bash
npm install -g trustmesh-mcp
```

## Configure

### Claude Desktop

Add to `claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "trustmesh": {
      "command": "npx",
      "args": ["-y", "trustmesh-mcp"],
      "env": {
        "TRUSTMESH_NODE_ID": "my-agent-001",
        "TRUSTMESH_STORAGE": "sqlite",
        "TRUSTMESH_DB_PATH": "/path/to/trust.db"
      }
    }
  }
}
```

### Cursor

Add to `.cursor/mcp.json`:

```json
{
  "mcpServers": {
    "trustmesh": {
      "command": "npx",
      "args": ["-y", "trustmesh-mcp"]
    }
  }
}
```

### Windsurf

Add to `.windsurf/mcp.json`:

```json
{
  "mcpServers": {
    "trustmesh": {
      "command": "npx",
      "args": ["-y", "trustmesh-mcp"]
    }
  }
}
```

## Environment Variables

| Variable | Default | Description |
|----------|---------|-------------|
| `TRUSTMESH_NODE_ID` | Random | Your agent's identity |
| `TRUSTMESH_STORAGE` | `memory` | `memory` or `sqlite` |
| `TRUSTMESH_DB_PATH` | `./trustmesh.db` | SQLite path (if using sqlite) |

## Usage Examples

**Before transacting with an unknown agent:**
```
> Check trust score for agent-paybot-42
→ Trust Score: 0.23 (🔴 LOW)
→ ⚠️ WARNING: This agent has very low trust. Proceed with extreme caution.
```

**After a successful API call:**
```
> Rate agent-weather-api 0.9 for delivered-api-data
→ Rating recorded. Trust graph updated.
```

**Batch check counterparties:**
```
> Check scores for [agent-a, agent-b, agent-c]
→ 🟢 agent-a: 0.85
→ 🟡 agent-b: 0.52
→ 🔴 agent-c: 0.18
```

## License

MIT