# Localize

> Localize is an open-source MCP server for ai models and llm tooling. Offload log reading, web research and test-gated coding to your local LLM. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted, Linux, macOS, Windows
- **Repository:** https://github.com/joblessjoe/localize
- **MCP Registry name:** io.github.JoblessJoe/localize
- **Categories:** [AI models and LLM tooling](https://ai-powered.com/categories/ai-and-llm), [developer tools](https://ai-powered.com/categories/developer-tools)
- **Listing updated:** 2026-11-18

## About

Localize targets a cost problem: reading a 5,000-line test log or a few docs pages burns the same frontier-model tokens as hard design work. It hands that bulk work to a model on your own machine and returns only the answer. In the README's example a 50 KB test log (about 19,100 tokens read locally) came back to Claude as roughly 130 tokens.

There are three working tools plus configuration. offload runs a noisy command or reads a large file and answers a question about it. research searches the web via a SearXNG URL, reads the pages in full and returns a cited answer that Claude never sees as pages. delegate is opt-in: the local model writes one file in an isolated git worktree and retries against a test Claude wrote first, feeding back parsed failures while keeping the test source hidden, and Claude gets a one-line verdict. A configure call shows and changes settings without restarting.

It works with Ollama, llama.cpp, LM Studio, vLLM, LocalAI or any OpenAI-compatible endpoint, with no dependencies beyond Node 18 and git. The author's benchmark of 120 delegate runs rates qwen3-coder 30B as the best default among the three models tried.

## Features

- **Offload noisy commands:** A local model reads logs and files and returns the answer.
- **Cited web research:** Searches and reads pages in full via SearXNG, returning a cited summary.
- **Test-gated delegation:** Writes one file in a git worktree until your test passes.
- **Parallel calls:** Several offloads or delegates run at once.
- **Any local backend:** Ollama, llama.cpp, LM Studio, vLLM and OpenAI-compatible servers.
- **Runtime configuration:** One configure call inspects and changes settings.

## Good for

- Find the one failing test in a huge log
- Answer questions from long documentation pages without filling context
- Delegate small well-specified helpers with a test written first
- Cut frontier-model token spend on routine reading

## Install

### Claude Code (Terminal)

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

### Claude Desktop (claude_desktop_config.json)

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

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

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

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

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

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

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

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

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

## Tools

- `offload`: Run a command or read a large file with the local model and return the answer.
- `research`: Search the web and return a cited answer from the full pages.
- `delegate`: Have the local model write one file until a given test passes.
- `configure`: Show or change configuration, backend and models.

## FAQ

### Which models work?

Any model served by Ollama, llama.cpp, LM Studio, vLLM, LocalAI or an OpenAI-compatible endpoint; CPU-only works but is slower.

### Is delegate on by default?

No, offload and research are the defaults and delegate is opt-in.

### What does it need to install?

Node 18 or newer, git and a running local model server.

### How does delegate avoid overwriting my work?

It works in an isolated worktree and does not apply the file if it changed meanwhile.

### Is Localize free?

Yes. Localize is open source, so you can use and self-host it for free.

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

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

### What are the best alternatives to Localize?

The closest alternatives to Localize are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.

## Where this comes from

- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Context7](https://ai-powered.com/mcp/context7): Context7 is a freemium MCP server for developer tools.
- [Chrome DevTools MCP](https://ai-powered.com/mcp/chrome-devtools-mcp-2): Chrome DevTools MCP is an open-source MCP server for browser automation.
- [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.
- [awesome-mcp-servers](https://ai-powered.com/mcp/awesome-mcp-servers-2): awesome-mcp-servers 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.

---
Source: https://ai-powered.com/mcp/localize
