IDE Chat Summarizer

You run an MCP server that converts AI chat conversations into organized, markdown summaries saved in your notes directory. This makes it easy to reference important discussions, decisions, and insights, with smart handling for large histories and code detection to preserve working solutions.

How to use

Power up your MCP client and connect to the chat summarizer server to convert conversations into structured summaries. You can summarize standard chats, handle very large histories with chunking, and review or delete existing summaries when needed. The server supports different summary styles, auto-generates timestamped filenames, and preserves code blocks and final solutions for easy reference.

How to install

Prerequisites you need to prepare before starting: the uv package manager and Python 3.13 or later.

Install dependencies by synchronizing MCP components.

uv sync

Starting the server with an MCP client

Choose your preferred start method based on your workflow. The server accepts several runtime options depending on whether you want to run it from the MCP Inspector web interface, directly from the command line, or via the MCP CLI.

Configuration and startup options

STDIO startup commands (local server) you can use directly from your MCP client or IDE integrations include the following. Each entry runs the MCP server as a local process via the uv runtime.

uv run mcp dev main.py

uv run python main.py

uv run mcp run main.py

Notes directory and environment

The server stores all generated summaries in your configured notes directory. A commonly used default is ~/Documents/ChatSummaries. You can set a custom directory via an environment variable or by editing the main configuration.

Client integration details

In your IDE, configure the MCP server with the appropriate command so the client can start the server automatically when you use the tools. For example, in a VS Code-like environment, point to the stdio start command and set an optional environment variable for a custom notes directory.

Additional usage and features

Use the summarize_chat tool to convert a chat into a markdown summary, with options for including the full history, creating separate full history files, and choosing a summary style such as detailed, brief, or bullet_points. The summarize_large_chat tool handles extremely large histories by chunking into manageable pieces and creating a master summary that links all chunks. You can list, read, and delete summaries as needed.

Smart code detection

The summarizer detects and preserves code from conversations. It captures code blocks, inline code, final solutions, and late-conversation code, and presents them with proper syntax highlighting and structured sections in the summary.

Directory information and organization

Summaries are saved to your notes directory. Filenames follow the pattern: chat_summary_YYYYMMDD_HHMMSS_[title].md for easy sorting and retrieval.

Troubleshooting and tips

If you need to clean up, use delete_summary to remove outdated or incorrect summaries. Review recent activity with list_summaries to see creation dates and sizes. When testing with MCP Inspector, you can open the web interface at http://localhost:6274 to try commands and observe behavior.

File organization and examples

Typical file names you might see include chat_summary_20240115_143022_Python_Optimization.md or chat_summary_20240115_150330_API_Design_Discussion.md.

Quick start for large histories

For very large conversations, you can summarize in standard mode with full history, or enable chunking to manage extremely large histories. The chunking approach creates a master summary plus individual chunk files with overlap for context.

Terminology and prompts

Prompts you may use include generating general, technical, meeting, or brainstorm summaries. Focus areas can include all content, decisions, action_items, and insights.

Client integration details beyond VS Code

If you use Cursor or Visual Studio, configure the MCP server similarly to ensure the server starts automatically when using the tools. You can also run the server directly for testing with MCP Inspector.

Security and safety considerations

Keep your notes directory protected and ensure that only authorized clients can access the summaries. When running in shared environments, consider restricting write permissions to the notes folder.

Project-wide notes

This server supports organized, searchable storage of AI conversation summaries with smart handling for large histories, code preservation, and flexible summary styles.

Available tools

summarize_chat

Summarize a chat conversation into a markdown summary with options for including full history, separate full history, and different styles.

summarize_large_chat

Handle extremely large histories by chunking into manageable pieces and creating a master summary that links all chunks.

list_summaries

List recent summaries with creation dates and sizes.

delete_summary

Delete a selected summary file with safety checks.

summary://filename

Read content of a specific summary file.

notes://directory

Get information about the configured notes directory.

create_summary_prompt

Generate customized prompts for different conversation types such as general, technical, meeting, or brainstorm.