claudex

Overview

Claudex is a full-stack web application designed for developers, QA engineers, and researchers to inspect, search, and analyze Claude Code conversation histories. It serves as an MCP server with 10 tools that give Claude Code persistent memory across sessions, storing coding knowledge like conventions, architecture decisions, and error patterns. The application includes enterprise-grade SQLite FTS5 full-text search, universal template support for all Claude Code versions, and a comprehensive analytics dashboard.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• search_messagessearch_messages(query, limit, project, session, role, before, after)

  Search through conversation messages using FTS5 full-text search

  Try:Find messages about authentication errorsTry:Search for all mentions of 'migration'Try:Find user messages from last week about database queries
• get_sessionget_session(project, session)

  Retrieve a complete conversation session by project and session ID

  Try:Get the session 'project-alpha-conversation-1'Try:Show me the full conversation about the payment featureTry:Retrieve session details for project X
• list_projectslist_projects()

  List all available Claude Code projects

  Try:What projects do I have?Try:List all my Claude conversationsTry:Show me available projects
• list_sessionslist_sessions(project, limit)

  List conversation sessions for a specific project

  Try:Show me all sessions for project-alphaTry:List recent conversations in my payments projectTry:What sessions do I have in project X?
• store_memorystore_memory(key, content, priority, confidence, ttl)

  Store structured coding knowledge with priority, confidence and expiration

  Try:Remember that we use PostgreSQL for all database operationsTry:Store our API authentication pattern as a best practiceTry:Record that we've decided to use React hooks instead of class components
• retrieve_memoryretrieve_memory(query, limit)

  Retrieve stored structured coding knowledge

  Try:What are our database conventions?Try:Remember our API authentication patternTry:Show me stored decisions about architecture
• update_memoryupdate_memory(key, content, priority, confidence, ttl)

  Update existing structured memory entry

  Try:Update our database conventions to include the new connection pooling approachTry:Modify our React patterns guidance to include hooksTry:Update the error handling pattern with the new middleware
• delete_memorydelete_memory(key)

  Remove a structured memory entry

  Try:Forget the old API authentication patternTry:Remove the deprecated React class component conventionTry:Delete the temporary database connection decision
• get_contextget_context(project, session, detail_level)

  Get project context with configurable detail level

  Try:Get minimal context for project-alphaTry:Show full context for the current sessionTry:Give me standard context about our database schema
• build_search_indexbuild_search_index()

  Rebuild the FTS5 search index for all conversations

  Try:Rebuild the search indexTry:Update search index with new conversationsTry:Refresh search database

Comparable tools

Installation

# Global installation
npm install -g @kunwarshah/claudex

# Then run:
claudex

claude mcp add --transport stdio claudex -- claudex-mcp

git clone https://github.com/kunwar-shah/claudex.git
cd claudex
npm run check:fix
npm run dev

FAQ

What file locations does Claudex scan?

Claudex automatically scans the ~/.claude/projects directory by default, but you can configure a custom path with --project-root flag.

Which Claude Code versions are supported?

Claudex supports universal template detection and parsing for all Claude Code versions (V1.x, V2-mixed, V2.0+).

Compare claudex with