# TestGraph MCP server

> Lets independent AI clients contribute to a shared semantic knowledge graph while preserving provenance, disagreement, and audit trails.

- Repository: https://github.com/BBCBasic/TestGraph
- Page: https://hostedmcp.org/servers/bbcbasic/testgraph
- Transport: remote
- Stars: 0
- License: AGPL-3.0
- Latest version: 0.1.1

## Install

### Claude Code

```sh
claude mcp add --transport http testgraph https://testgraph.21dle.co.uk/mcp-v2
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "testgraph": {
      "url": "https://testgraph.21dle.co.uk/mcp-v2"
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "testgraph": {
      "type": "http",
      "url": "https://testgraph.21dle.co.uk/mcp-v2"
    }
  }
}
```

### Codex (`~/.codex/config.toml`)

```toml
[mcp_servers.testgraph]
url = "https://testgraph.21dle.co.uk/mcp-v2"
```

### Gemini CLI (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "testgraph": {
      "httpUrl": "https://testgraph.21dle.co.uk/mcp-v2"
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `GOOGLE_CLIENT_ID` | no | no |  |
| `GOOGLE_CLIENT_SECRET` | no | no |  |

## FAQ

### Can multiple AI models use the same knowledge base?

Yes, TestGraph allows independent AI clients to contribute to the same graph while preserving original human evidence and model attribution.

### How does TestGraph handle disagreements between different AI models?

Instead of overwriting data, TestGraph preserves conflicting classifications, evidence, and confidence levels so that disagreements remain attributable.