# 🔧 Connecting Your AI to Taskr

## 2.8 Other AI Tools

If you're using something else that supports MCP, here's the universal setup.

### Interactive Tools (with a browser)

Most modern AI tools can connect to a **remote MCP server** over the web. Point yours at Taskr's URL—no install, no keys. You'll sign in through your browser on first connect:

```json
{
  "taskr": {
    "url": "https://www.taskr.one/api/mcp"
  }
}
```

Some tools use a slightly different key for the address (e.g. `serverUrl`) or want a `"type": "http"` field—check your tool's MCP docs. When the tool connects, it opens your browser to sign in and approve.

### Headless Tools (servers, CI, no browser)

If your tool runs on a server with no browser, send your **API key** as a request header instead of doing the browser sign-in:

```
x-user-api-key: YOUR_API_KEY_HERE
```

Get your API key from Taskr (profile → **API Keys**). Point the tool at `https://www.taskr.one/api/mcp` and add the header above. To pin a project, also send `x-project-id: PR_...` (optional).

### Legacy: the local `npx` bridge

If your tool only supports local (stdio) MCP servers and can't reach a URL, you can still run the local connector (requires Node.js):

```json
{
  "taskr": {
    "command": "npx",
    "args": ["-y", "taskr-mcp-client@latest"]
  }
}
```

It signs you in through your browser on first run (or run `npx taskr-mcp-client@latest auth` once). For headless use, add an `env` block with `MCP_USER_API_KEY`.

### Where to Put It

Look for:
- A file ending in `.json` with "mcp" in the name
- Settings or preferences mentioning "MCP" or "Model Context Protocol"
- Documentation about "external tools" or "integrations"

---