# Paw Pulse

> Paw Pulse is a free MCP server for ai models and llm tooling. Heartbeat and tokens-per-second monitoring for home Ollama, llama.cpp and vLLM servers. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** Web, API, Linux, macOS
- **Website:** https://pulse.dougalpaw.com
- **Repository:** https://github.com/dougal-paw/pawpulse
- **MCP Registry name:** com.dougalpaw/pawpulse
- **Categories:** AI models and LLM tooling, developer tools
- **Listing updated:** 2026-11-14

## About

Most uptime monitors poll from outside, which means opening a firewall port on your LLM box. Paw Pulse reverses it: a small shell script on your machine, run from cron or a systemd timer, sends a heartbeat out every few minutes. If heartbeats stop for longer than the grace period you set, the monitor goes down and you are alerted. The server never accepts an inbound connection.

The script probes Ollama or any OpenAI-compatible /v1/models endpoint, reports installed models, and can optionally time a tiny generation to estimate tokens per second. On Ollama with jq installed it also reports how much of the loaded model sits in GPU memory, so you can be told when it falls back to CPU, or when a different model than expected is loaded. With LLM_KIND=none it works as a plain heartbeat for a NAS or backup job.

Alerts arrive when a monitor goes down, slows below your threshold and recovers, once each. Discord, Slack, Teams, Google Chat and Telegram webhook URLs are recognised, with generic JSON for others. Monitors are anonymous, identified by an id plus ingest and manage tokens, and the service says it stores no IPs or personal data. The MCP server lets an agent create a monitor, fetch setup commands and check status. The site says it is free during beta and operated by an AI agent.

## Features

- **Push-based heartbeats:** A script sends heartbeats out, so no inbound ports are needed.
- **Speed alerts:** Get notified when tokens per second fall below your threshold.
- **GPU fallback check:** Detect when Ollama moves a model from GPU to CPU.
- **Webhook alerts:** Discord, Slack, Teams, Google Chat, Telegram or generic JSON.
- **Public status:** Status page, JSON endpoint and SVG badge per monitor.
- **Anonymous monitors:** No accounts; each monitor uses its own tokens.

## Good for

- Learn quickly when a home Ollama box has crashed
- Spot slow generation before users complain
- Catch a silent switch from GPU to CPU
- Monitor a NAS or cron backup with a plain heartbeat
- Let an agent set up monitoring from a chat

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http pawpulse https://pulse.dougalpaw.com/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://pulse.dougalpaw.com/mcp
```

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

```json
{
  "mcpServers": {
    "pawpulse": {
      "url": "https://pulse.dougalpaw.com/mcp"
    }
  }
}
```

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

```json
{
  "servers": {
    "pawpulse": {
      "type": "http",
      "url": "https://pulse.dougalpaw.com/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "pawpulse": {
      "serverUrl": "https://pulse.dougalpaw.com/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "pawpulse": {
      "type": "http",
      "url": "https://pulse.dougalpaw.com/mcp"
    }
  }
}
```

## Tools

- `create_monitor`: Create a monitor.
- `setup_instructions`: Return setup commands for a monitor.
- `get_status`: Return a monitor's current status.
- `list_heartbeats`: List recent heartbeats.
- `update_monitor`: Update a monitor's settings.
- `delete_monitor`: Delete a monitor.

## FAQ

### Do I need to open a firewall port?

No. The client pushes heartbeats outward.

### What does it cost?

The site says it is free during beta.

### Which servers can it watch?

Ollama, llama.cpp, vLLM or any OpenAI-compatible endpoint, or none with LLM_KIND=none.

### Is the tokens-per-second probe on by default?

No. It is off by default because it uses some compute on your machine.

### Is Paw Pulse free?

Yes. Paw Pulse is free to use.

### How do I add Paw Pulse to Claude Code?

Run this command in your terminal: claude mcp add --transport http pawpulse https://pulse.dougalpaw.com/mcp

### What are the best alternatives to Paw Pulse?

The closest alternatives to Paw Pulse are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/paw-pulse
