Sherlog MCP Server

Sherlog MCP Server is an MCP server that provides a persistent IPython shell per session for data analysis, log processing, and multi-agent workflows. It connects LLM clients like Claude Desktop to containerized execution environments, serving developers, data engineers, and data analysts who need stateful computation. By keeping tool outputs inside the IPython namespace as variables and DataFrames, the server helps LLMs avoid context window exhaustion when dealing with large payloads. Users interact with the environment primarily through core tools like execute_python_code and call_cli, alongside supporting utilities for package installation, code retrieval with Tree-sitter, and shell variable introspection. The server also functions as an MCP proxy, allowing external MCP servers to execute inside the same IPython runtime so their results automatically become manipulable DataFrames. It features persistent storage across restarts, supports multi-session lifecycle management for up to four concurrent sessions, and offers built-in Google OAuth 2.0 integration for accessing Google Workspace resources.

Category: Data & Analytics

Tags: data analysis, ipython, jupyter, log processing, python

Visit Sherlog MCP Server

How to install and configure Sherlog MCP Server

1. Ensure Docker Desktop is installed and running on your host machine. 2. Clone the repository from https://github.com/GetSherlog/Sherlog-MCP or prepare the container deployment. 3. Configure any required environment variables such as MCP_MAX_SESSIONS, MCP_AUTO_RESET_THRESHOLD, or proxy settings via EXTERNAL_MCPS_JSON. 4. Set up Google Workspace access using GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET if OAuth features are needed. 5. Run the server container locally or deploy it to Railway using persistent volume mounts on /app/data. 6. Connect your MCP client, such as Claude Desktop, via HTTP transport by following the repository's remote connection setup instructions at https://github.com/GetSherlog/Sherlog-MCP.

What you can do with Sherlog MCP Server

• Inspecting and filtering massive log files inside an IPython DataFrame without flooding the model context window. * Executing iterative Python workflows and data transformations across persistent session states that survive container reboots. * Running command-line tools and scripts via the composable call_cli interface within isolated container environments. * Proxying external MCP servers like PostgreSQL or Filesystem to directly capture structured query outputs as DataFrames. * Developing and debugging Android applications using pre-installed Android SDK, ADB, and BrowserStack CLI containers.

Key facts

• https://github.com/GetSherlog/Sherlog-MCP
• Data & Analytics, Developer Tools & Code Intelligence
• data analysis, ipython, jupyter, log processing, python

Part of MCP Servers

What can Sherlog MCP Server do?

Sherlog MCP Server provides isolated, persistent IPython shells for LLMs to run Python code and CLI commands. Instead of dumping large outputs directly into the chat context, it stores operation results as DataFrames and shell variables. It also acts as an MCP proxy to run third-party MCP servers inside the same execution environment.

Which MCP clients work with Sherlog MCP Server?

It is designed to work with Claude Desktop and other Model Context Protocol clients capable of connecting to remote or local servers via HTTP transport.

How does Sherlog MCP Server handle external MCP servers?

External MCP servers can be configured using the EXTERNAL_MCPS_JSON environment variable. When an external tool is triggered, its outputs are automatically converted into DataFrames within the shared IPython namespace, making them immediately accessible for subsequent Python code execution.

Is Sherlog MCP Server open source?

Yes, Sherlog MCP Server is open source and licensed under the Apache License 2.0. You can review the source code and license directly in its official GitHub repository.

How are sessions persisted across restarts?

The server stores session files and metadata in the /app/data directory. When deployed with persistent storage, such as a Railway volume, user shell sessions, imports, and variables are automatically preserved across container restarts.