# Llmxray

> Llmxray is a free MCP server for ai models and llm tooling. See what Ollama is really doing with your prompts, with plain-language fixes. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** self-hosted, CLI, Web
- **Website:** https://lognebudo.github.io/llmxray/
- **Repository:** https://github.com/lognebudo/llmxray
- **MCP Registry name:** io.github.LogneBudo/llmxray
- **Categories:** AI models and LLM tooling, developer tools
- **Listing updated:** 2026-11-14

## About

LLMxRay sits between your tools and a local Ollama on a second port, 11435, forwards everything and records counts and timings. The MCP server lets an agent read that record and ask what went wrong with its own recent calls. It is read-only, runs over stdio from the llmxray npm package, and reads from the running LLMxRay server plus the Ollama daemon for its list_local_models tool.

Findings are written as sentences with the fix: a chat that outgrew the context and lost its oldest messages (raise num_ctx), a changing timestamp near the top of the prompt that breaks the cache (move it to the end), two tools evicting each other's model on one GPU (keep_alive and num_ctx settings), and agents that start a fresh conversation at every step. Only counts and timings are stored, on your machine for 30 days; prompt text is opt-in.

The site claims it is free and local, with no account or key, and turns can be sent on as OpenTelemetry spans.

## Features

- **Plain-sentence findings:** Each diagnosis links to the turn that proves it.
- **Cache break detection:** Shows where a prompt stopped matching the previous one.
- **Silent truncation:** Reveals when Ollama cut context to fit.
- **GPU contention:** Names which tool evicted which model and what it cost.
- **Local and private:** Counts and timings stay on your machine for 30 days.
- **OpenTelemetry export:** Send turns on to Jaeger, Phoenix, Langfuse or Grafana.

## Good for

- Ask an agent why its last Ollama calls were slow
- Find out whether a prompt was truncated by the context window
- Spot a timestamp that keeps invalidating the cache
- Work out which tool is hogging a shared GPU
- List models available on the local Ollama

## Install

### Claude Code (Terminal)

```sh
claude mcp add llmxray -- npx -y llmxray
```

### Claude Desktop (claude_desktop_config.json)

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

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

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

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

```json
{
  "servers": {
    "llmxray": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "llmxray"
      ]
    }
  }
}
```

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

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

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

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

## Tools

- `list_local_models`: Lists models on the local Ollama daemon.

## FAQ

### Does it need an account or API key?

No. It is described as free and local with no cloud.

### How do I point my tools at it?

Use port 11435 instead of Ollama's 11434.

### Does it store my prompts?

Only counts and timings by default; prompt text is an opt-in written to a local file.

### How do I start it?

Run npx llmxray or the Docker image, then add the llmxray-mcp server to your client.

### Is Llmxray free?

Yes. Llmxray is free to use.

### How do I add Llmxray to Claude Code?

Run this command in your terminal: claude mcp add llmxray -- npx -y llmxray

### What are the best alternatives to Llmxray?

The closest alternatives to Llmxray 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/llmxray
