Claude Code Session Search MCP Server

An MCP (Model Context Protocol) server that provides tools for searching and analyzing Claude Code conversation history.

Features

• List Projects: View all Claude Code projects with session counts
• List Sessions: Browse sessions for specific projects
• List Recent Sessions: Find recent conversations across all projects
• Analyze Sessions: Extract and analyze messages with role filtering
• Search Conversations: Search for specific terms with context windows and time ranges
• Get Message Details: Retrieve full content for specific messages
• Summarize Conversations: AI-powered summarization of daily conversations

Installation

1. Install dependencies:

uv sync

2. Run the server:

uv run python server.py

3. Add to Claude Code MCP config (~/.config/claude/mcp.json):

{
  "servers": {
    "cc-session-search": {
      "command": ["uv", "run", "python", "server.py"],
      "cwd": "/path/to/cc-session-search"
    }
  }
}

Requirements

• Standard Claude Code installation (searches ~/.claude/projects/)
• Python 3.13+
• MCP 1.2.0+

Usage

The server provides the following tools:

list_projects()

Lists all Claude Code projects with session counts and recent activity.

list_sessions(project_name, days_back=7)

Lists sessions for a specific project within the specified time range.

list_recent_sessions(days_back=1, project_filter=None)

Lists recent sessions across all projects.

analyze_sessions(days_back=1, role_filter="both", include_tools=False, project_filter=None)

Extracts and analyzes messages from sessions with filtering options.

search_conversations(query, days_back=2, context_window=1, case_sensitive=False, project_filter=None)

Searches conversations for specific terms with context windows.

get_message_details(session_id, message_indices)

Retrieves full content for specific messages by session ID and indices.

Development

The server is built using the official MCP Python SDK with low-level Server class for maximum control.

Key features:

• Efficient response handling with content truncation
• Metadata-first approach to minimize token usage
• Support for date ranges and filtering
• Cross-project search capabilities

License

MIT