claude-historian-mcp and claude-server
These are **complements**: the historian tool provides search/retrieval of conversation history while the server tool provides the persistent storage and hierarchical context management infrastructure that the historian queries against.
About claude-historian-mcp
Vvkmnn/claude-historian-mcp
📜 An MCP server for conversation history search and retrieval in Claude Code
Implements an MCP server that indexes Claude Code conversation history locally using search algorithms—no external dependencies or databases—enabling semantic retrieval across 11 scopes (conversations, errors, files, tools, plans, config, memories, and more). Provides two core tools: `search` with filtering by project, timeframe, and detail level, and `inspect` for session summaries with focus options on files, tools, or solutions. Integrates with Claude Code via stdio transport and optionally pairs with a companion skill and plugin ecosystem for proactive history checks before web searches and error handling.
About claude-server
davidteren/claude-server
Claude Server is an MCP implementation that enhances Claude's capabilities by providing sophisticated context management across sessions, enabling persistent knowledge organization through hierarchical project contexts and continuous conversation threads stored in a well-structured ~/.claude directory.
Implemented as an MCP server using stdio transport to manage hierarchical project contexts and conversation threads via JSON files in `~/.claude/`, enabling Claude to maintain stateful knowledge across disconnected sessions. The architecture stores conversation metadata and project hierarchies locally, allowing Claude Code to query persistent context without external dependencies. However, the project is archived as Claude Code's native auto-memory and CLAUDE.md features now provide equivalent functionality natively.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work