codex-agent-mem

Overview

codex-agent-mem is a portable, auditable, local-first memory layer for MCP-compatible AI agents and coding workflows. It stores all data locally using SQLite database with FTS5 indexing, compressing operational context into smaller packs to reduce token usage and improve continuity. The MCP server supports read-only mode, optional daemon/stdio bridge, and works with multiple AI environments including Codex CLI/Desktop, Claude Code, Gemini CLI, Qwen/DeepSeek/Ollama, and other local agent stacks. All processing happens locally without sending memory or telemetry to external servers.

Try asking AI

After installing, here are 3 things you can ask your AI assistant:

• mem_context_pack

  Compacts repeated session context into smaller working packs for token efficiency

  Try:Get the project context packTry:Retrieve compact continuity for this session
• mem_search

  Searches through stored memory and project data

  Try:Search for previous work related to authenticationTry:Find notes about API integration
• mem_note_create

  Creates and indexes manual notes in the local memory store

  Try:Create a note about this bug fixTry:Add a reminder for the team meeting
• mem_snapshot_create

  Creates a snapshot of the current project state with provenance tracking

  Try:Take a snapshot before making breaking changesTry:Create a checkpoint for the current feature
• mem_open_work

  Tracks and manages open work items across sessions

  Try:List all open work itemsTry:Add a new task to the work queue
• mem_completion_check

  Checks if all defined criteria for work completion have been met

  Try:Verify if the current task meets Definition of DoneTry:Check if all objectives are completed
• mem_session_list

  Lists recent memory sessions for session-aware retrieval

  Try:Show me recent sessionsTry:List all previous work sessions
• mem_scope_resolve

  Ranks persisted memory lanes from explicit thread/path hints

  Try:Resolve scope for the authentication moduleTry:Get context for the payment processing feature
• mem_bootstrap_context

  Avoids project-wide startup packs for ambiguous containers

  Try:Get bootstrap context for the new containerTry:Initialize context for this specific directory
• mem_health_runtime

  Checks runtime health and diagnostics

  Try:Check server health statusTry:Get runtime diagnostics

Comparable tools

Installation

pip install codex-agent-mem

{
  "mcpServers": {
    "codex-agent-mem": {
      "command": "codex-agent-mem",
      "args": ["--read-only", "--profile", "minimal"]
    }
  }
}

Compare codex-agent-mem with