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

# MCP Server

> Connect Claude Code, Cursor, or Windsurf to LogClaw incidents, logs, and anomalies via the Model Context Protocol.

import { Card, CardGroup, CodeGroup, Info, Tip } from "mintlify/components";

Connect your AI coding tools to LogClaw incidents, logs, and anomalies via the [Model Context Protocol](https://modelcontextprotocol.io). Works with **Claude Code**, **Cursor**, **Windsurf**, **Claude.ai**, and any MCP-compatible client.

<CardGroup cols={2}>
  <Card title="Remote (Cloud)" icon="cloud" href="#remote-cloud">
    Connect via URL — no installation needed
  </Card>

  <Card title="Local (npm)" icon="npm" href="#local-npm">
    Install locally via `npx logclaw-mcp-server`
  </Card>
</CardGroup>

***

## Remote (Cloud)

Connect directly via URL — no installation required. LogClaw's remote MCP server uses **OAuth 2.1** and **Streamable HTTP** transport.

```
https://mcp.logclaw.ai
```

<Info>
  When you connect, you'll be prompted to enter your LogClaw API key. The key is encrypted and never exposed to the MCP client.
</Info>

### Claude Desktop / Claude.ai

Add LogClaw as a remote MCP server:

1. Open **Settings → MCP Servers → Add Server**
2. Enter the URL: `https://mcp.logclaw.ai`
3. Complete the OAuth flow by entering your API key

### Any MCP Client (URL-based)

For any MCP client that supports remote servers, use:

| Setting        | Value                    |
| -------------- | ------------------------ |
| **Server URL** | `https://mcp.logclaw.ai` |
| **Transport**  | Streamable HTTP          |
| **Auth**       | OAuth 2.1 (automatic)    |

***

## Local (npm)

Install and run the MCP server locally via stdio transport:

```bash theme={null}
npx logclaw-mcp-server
```

The server communicates over **stdio** using the Model Context Protocol. Your AI tool launches it automatically — you just need to add the config once.

### Setup

<CodeGroup>
  ```json Claude Code (~/.claude.json) theme={null}
  {
    "mcpServers": {
      "logclaw": {
        "command": "npx",
        "args": ["-y", "logclaw-mcp-server"],
        "env": {
          "LOGCLAW_ENDPOINT": "https://ticket.logclaw.ai",
          "LOGCLAW_API_KEY": "lc_proj_..."
        }
      }
    }
  }
  ```

  ```json Cursor (.cursor/mcp.json) theme={null}
  {
    "mcpServers": {
      "logclaw": {
        "command": "npx",
        "args": ["-y", "logclaw-mcp-server"],
        "env": {
          "LOGCLAW_ENDPOINT": "https://ticket.logclaw.ai",
          "LOGCLAW_API_KEY": "lc_proj_..."
        }
      }
    }
  }
  ```

  ```json Windsurf (.windsurf/mcp.json) theme={null}
  {
    "mcpServers": {
      "logclaw": {
        "command": "npx",
        "args": ["-y", "logclaw-mcp-server"],
        "env": {
          "LOGCLAW_ENDPOINT": "https://ticket.logclaw.ai",
          "LOGCLAW_API_KEY": "lc_proj_..."
        }
      }
    }
  }
  ```
</CodeGroup>

***

## Environment Variables

| Variable           | Required | Description                                                |
| ------------------ | -------- | ---------------------------------------------------------- |
| `LOGCLAW_API_KEY`  | Yes      | Your LogClaw project API key (starts with `lc_proj_`)      |
| `LOGCLAW_ENDPOINT` | No       | Auth proxy endpoint (default: `https://ticket.logclaw.ai`) |

Get your API key from the [LogClaw Console](https://console.logclaw.ai) under **Settings → API Keys**.

***

## Available Tools

The MCP server exposes 8 tools that your AI assistant can call:

| Tool                       | Description                                                                                     |
| -------------------------- | ----------------------------------------------------------------------------------------------- |
| `logclaw_list_incidents`   | List and filter incidents by severity, state, service, or search query                          |
| `logclaw_get_incident`     | Full incident details — root cause, causal chain, evidence logs, traces, timeline, blast radius |
| `logclaw_update_incident`  | Transition incident state (acknowledge, investigate, mitigate, resolve) or add a note           |
| `logclaw_search_logs`      | Search raw logs by service, level, time range, and query                                        |
| `logclaw_get_anomalies`    | Get recent anomaly detections (Z-score analysis on error rates)                                 |
| `logclaw_service_health`   | Check LogClaw pipeline health status and latency                                                |
| `logclaw_bulk_update`      | Update multiple incidents at once (acknowledge, investigate, mitigate, or resolve)              |
| `logclaw_export_incidents` | Export incidents in CSV, markdown table, or JSON format                                         |

***

## Example Prompts

Once connected, try asking your AI assistant:

<CardGroup cols={2}>
  <Card title="Triage incidents" icon="fire">
    "What incidents are open right now?"
  </Card>

  <Card title="Root cause analysis" icon="magnifying-glass">
    "Show me the root cause for TICK-0037"
  </Card>

  <Card title="Search logs" icon="scroll">
    "Search for ERROR logs from auth-service in the last 30 minutes"
  </Card>

  <Card title="Query anomalies" icon="chart-line">
    "Are there any critical anomalies?"
  </Card>

  <Card title="Update incidents" icon="check">
    "Acknowledge TICK-0042 and add a note that I'm looking into it"
  </Card>

  <Card title="Health check" icon="heart-pulse">
    "Is the LogClaw pipeline healthy?"
  </Card>

  <Card title="Bulk update" icon="list-check">
    "Acknowledge TICK-0037, TICK-0038, and TICK-0039"
  </Card>

  <Card title="Export incidents" icon="file-export">
    "Export all critical incidents as a markdown table"
  </Card>
</CardGroup>

***

## Self-Hosted

Running LogClaw on your own infrastructure? Point `LOGCLAW_ENDPOINT` to your auth proxy:

```bash theme={null}
LOGCLAW_ENDPOINT=https://logclaw.internal.company.com \
  LOGCLAW_API_KEY=lc_proj_... \
  npx logclaw-mcp-server
```

<Tip>
  **Air-gapped?** Install the package from a private registry or copy the `dist/` folder directly. The MCP server has zero runtime network dependencies beyond reaching your auth proxy.
</Tip>

***

## Source Code

<CardGroup cols={2}>
  <Card title="GitHub" icon="github" href="https://github.com/logclaw/logclaw/tree/main/apps/logclaw-mcp-server">
    View source code and contribute
  </Card>

  <Card title="npm" icon="npm" href="https://www.npmjs.com/package/logclaw-mcp-server">
    logclaw-mcp-server on npm
  </Card>
</CardGroup>