A persistent span of identity, memory, and conversation for LLMs and agents
LLMs don't remember across sessions. Platform memories are locked to single tools. Your identity resets every conversation.
memspan bridges this gap by extracting, organizing, and loading your personal identity, saved memories, and conversation history into Claude Code and other LLM interfaces. It provides a portable, tool-agnostic approach to maintaining continuity across AI assistant sessions.
memspan is independent of Claude Code's own memory features and doesn't interfere with them.
A three-tier memory model. Core identity (up to ~10KB), project memory (~10-50KB), and historical archive.
Core Identity (up to ~10KB): Always-available personal context
Project/Framework Memory (~10-50KB per domain): Session-selectable deep context
Historical Archive: Indexed, retrieved on-demand
File-Based
No databases, no servers. Just files you control. Your memory lives in plain Markdown and JSON that you own.
Portable
Works across different LLM tools and platforms. Not locked to ChatGPT, Claude, or any single service.
Selective Loading
Load only what you need for each session. Identity only, project context, or full memory archive.
Identity Extraction
Scripts and prompts to extract your personal identity from ChatGPT conversations into structured files.
Memory Export
Tools to export and structure ChatGPT saved memories. Copy or generate structured exports easily.
Context Loading
A wrapper script to dynamically load identity, memories, and project context into Claude Code sessions.
Proactive Memory Saving
Claude automatically detects notable information (goals, preferences, insights) and offers to save it. When confirmed, memories are stored for future sessions.
Selective loading for every scenario. From quick questions to deep project work.
cc-memspan --identity Load just your identity. Claude knows who you are, your preferences, and your communication style. Ask 'who am I?' and Claude responds with a personalized summary based on your identity context.
cc-memspan --memories Load your saved memories from ChatGPT and Claude. Recall past conversations and learned preferences.
Automatic during sessions Claude automatically detects notable information (goals, preferences, insights) and offers to save it. When confirmed, memories are stored for future sessions. Works in ask-first mode by default.
cc-memspan --full project-name Load identity + memories + project context. Everything Claude needs to work on a specific project.
claude "your question" For generic questions, skip the context. No loading overhead when you don't need memory.
Built for personal identity continuity across AI tools. Not for agents—for you.
Unlike frameworks built for agents, Memspan prioritizes personal identity preservation over isolated fact recall. Your identity, preferences, and communication style persist across tools.
Filesystem is the source of truth—human-readable, version-controllable, portable. Optional enhancements (MCP, database index, vector search) available when needed, but files always remain primary.
Works with any LLM interface. Not locked to ChatGPT, Claude, or any single service. Your memory travels with you.
You decide what context loads per session. No silent summarization or automatic injection. Full transparency and control.
| Framework | Target | Storage | Infrastructure | Identity Focus | Zero Setup |
|---|---|---|---|---|---|
| Memspan | People (personal continuity) | Files (default) | None (optional) | ||
| Letta/MemGPT | Agents (application memory) | Database + services | Required | ||
| Mem0 | Apps & agents | Database + vectors | Required | ||
| BasicMemory | Knowledge workers | Files + database | SQLite minimal |