Enables AI agents and MCP clients to search, read, and reference OpenAI Codex CLI session history.
Sponsored
This MCP server bridges the gap between disconnected AI agent histories, allowing any Model Context Protocol (MCP) client—such as Claude Code, Cursor, or Windsurf—to access your local OpenAI Codex CLI session data. It solves the problem of needing to re-explain or copy-paste context between agents by providing read-only access to past Codex conversations. The solution operates entirely locally, requiring no API keys, and processes both session metadata and full conversation logs to reconstruct detailed interactions, ensuring your work across different AI tools is unified and accessible.
主な機能
01Search and list Codex sessions by keyword with metadata
02Read full Codex conversations, including user messages, agent responses, and tool calls
03No API keys or external data transfer required; operates entirely locally
040 GitHub stars
05Provides read-only access to local Codex session history (state.sqlite and session JSONL files)
06Compatible with any MCP client for seamless integration
ユースケース
01Allow Claude Code or other MCP clients to continue work from a previous Codex CLI session
02Enable AI agents to reference and learn from past discussions and actions taken in Codex CLI
03Unify conversational context across multiple AI tools, eliminating the need for manual context transfer
