RLM Tools

Provides a server-side sandbox for AI agents to explore code with persistent memory, reducing context size while enabling deep analysis.

python

1

GitHub Stars

python

Language

2 months ago

First Indexed

3 weeks ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "stefanoshea-rlm-tools": {
      "command": "uvx",
      "args": [
        "rlm-tools"
      ],
      "env": {
        "RLM_SUB_MODEL": "claude-haiku-4-5-20251001",
        "RLM_MAX_SESSIONS": "5",
        "ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>",
        "RLM_SESSION_TIMEOUT": "10"
      }
    }
  }
}

RLM Tools provides a persistent server-side sandbox for your AI coding agent to explore code without bloating the context window. Data stays on the server, while only concise conclusions are returned to you, enabling deeper exploration with reduced token usage.

How to use

You connect to the RLM Tools MCP server from your MCP client (Claude Code, Codex, Cursor, or another compatible client) and start a sandboxed Python session that keeps data in memory on the server. Use the sandbox to grep, read files, or inspect a project’s structure, and pull back only compact outputs to guide your reasoning. Build incremental understanding by performing searches and reads inside the sandbox, then retrieve results as needed.

How to install

Prerequisites: you need an MCP client that supports standard MCP servers and the UVX runtime installed on your machine.

Choose one of the explicit MCP startup methods shown below. You can install using a quick one-liner or run from source as described.

Option A — Quick one-line install (Claude Code users)

claude mcp add rlm-tools -- uvx rlm-tools

Option B — Quick one-line install (Codex users)

codex mcp add rlm-tools -- uvx rlm-tools

Option C — Use the MCP server configuration directly (recommended if you manage MCP servers via JSON config)

{
  "mcpServers": {
    "rlm-tools": {
      "command": "uvx",
      "args": ["rlm-tools"]
    }
  }
}

Available tools

grep

Search for a pattern across files and return matches, with optional summary helpers like grep_summary and grep_read to tailor outputs.

read_file

Read a single file into a variable for processing within the sandbox.

read_files

Read multiple files into memory for batch analysis.

glob_files

Find files by pattern and iterate results in Python for selective processing.

tree

Inspect directory structure up to a specified depth to understand project layout.

llm_query

Perform semantic analysis inside the sandbox using an LLM API key (optional) for deeper insights.