Getting Started

Follow these steps to extract your identity, export memories, and start using memspan with Claude Code.

For complete documentation, see the full README on GitHub

Quick Start

1

Extract Your Identity

Use the prompts in identity-archive/ to extract your identity from ChatGPT conversations. This creates a structured identity file that Claude can use to understand who you are.

Location in repository:

identity-archive/
2

Export ChatGPT Memories

Follow the instructions in export-chatgpt-memories/README.md to export your ChatGPT memories. You can either copy from ChatGPT Settings → Memory or use the export prompt to generate structured exports.

Location in repository:

export-chatgpt-memories/
3

Correlate Project Conversations

To export your conversations (and optionally correlate with projects), use the scripts in export-chatgpt-conversations/:

Export projects metadata (if you use ChatGPT Projects):

python3 export-chatgpt-conversations/chatgpt_projects_dump.py --curl-file curl.txt

Correlate conversations with projects:

python3 export-chatgpt-conversations/chatgpt_project_conversations.py export
4

Load Context in Claude Code

Use the cc-memspan wrapper to load context into Claude Code. Here are the recommended usage patterns:

Minimal Load (Recommended Starting Point)

Core identity and memories:

bash claude-memory/bin/cc-memspan --identity --memories

Lightweight Project Awareness

Add projects index (lightweight - just project names/metadata). Requires projects.json from export-chatgpt-conversations:

bash claude-memory/bin/cc-memspan --identity --memories --projects-index

Full Project Context (Higher Token Usage)

Load specific project with full context (context.md, conversations.json). Use ad-hoc for long-running projects with essential ChatGPT conversations:

bash claude-memory/bin/cc-memspan --project project-name

Note: Loading full project context includes conversation history and can significantly increase token usage. Use --projects-index for lightweight project awareness, and reserve --project for when you need deep context from specific long-running projects.

Usage Examples

Basic Session with Identity

Load just your identity to give Claude context about who you are, your preferences, and your communication style.

bash claude-memory/bin/cc-memspan --identity

After loading identity, you can ask "who am I?" and Claude will respond with a summary based on your identity context.

Project-Specific Work

Load identity + memories + project context for full context on a specific project.

bash claude-memory/bin/cc-memspan --full your_project_name

This loads everything Claude needs to work on a specific project with full context.

Proactive Memory Saving

When you share notable information (goals, preferences, insights), Claude will proactively suggest saving it to your memory system. When you confirm, Claude creates a new entry file and makes the memory available for future sessions.

How It Works

Context Loading

The cc-memspan script reads selected context files (identity, memories, projects), combines them into a system prompt, and launches Claude Code with --append-system-prompt. This means the context is sent with every message, so Claude has access to your identity and memories throughout the session.

Architecture Overview

The following diagram shows how data flows from ChatGPT through the export process into the memspan file system, and then into Claude Code sessions:

graph TB classDef chatgptNode fill:#1e40af,stroke:#3b82f6,stroke-width:2px,color:#e2e8f0 classDef memspanNode fill:#92400e,stroke:#f59e0b,stroke-width:2px,color:#fef3c7 classDef claudeNode fill:#7c3aed,stroke:#a855f7,stroke-width:2px,color:#f3e8ff classDef exportNode fill:#1e293b,stroke:#475569,stroke-width:2px,color:#e2e8f0 subgraph ChatGPT["ChatGPT Platform"] CH_ID[Identity Data] CH_MEM[Saved Memories] CH_CONV[Conversations & Projects] end subgraph Export["Export Process"] EXP_ID[Identity Extraction
Manual Prompt] EXP_MEM[Memory Export
Settings UI or Prompt] EXP_CONV[Conversation Export
Python Scripts] end subgraph Memspan["memspan File System"] FS_ID[memory/identity/
core-identity.json] FS_MEM[memory/chatgpt/
memories_export.md] FS_PROJ[memory/projects/
projects.json
context.md] FS_CLAUDE[memory/claude/
index.json
entries/*.md] end subgraph CC["cc-memspan Script"] CC_LOAD[Load Context Files] CC_COMBINE[Combine into
System Prompt] end subgraph Claude["Claude Code Session"] CL_SESSION[Active Session] CL_SAVE[Proactive Memory Saving] end CH_ID -->|Extract| EXP_ID CH_MEM -->|Export| EXP_MEM CH_CONV -->|Export| EXP_CONV EXP_ID -->|Save| FS_ID EXP_MEM -->|Save| FS_MEM EXP_CONV -->|Save| FS_PROJ FS_ID -->|Load| CC_LOAD FS_MEM -->|Load| CC_LOAD FS_PROJ -->|Load| CC_LOAD FS_CLAUDE -->|Load| CC_LOAD CC_LOAD --> CC_COMBINE CC_COMBINE -->|Launch with| CL_SESSION CL_SESSION -->|Detects Notable Info| CL_SAVE CL_SAVE -->|Saves to| FS_CLAUDE FS_CLAUDE -->|Available for| CC_LOAD class CH_ID,CH_MEM,CH_CONV chatgptNode class FS_ID,FS_MEM,FS_PROJ memspanNode class FS_CLAUDE,CL_SESSION,CL_SAVE claudeNode class EXP_ID,EXP_MEM,EXP_CONV,CC_LOAD,CC_COMBINE exportNode

Memory Precedence

When information conflicts:

  • Claude memories (in memory/claude/) take precedence over ChatGPT memories
  • Newer entries take precedence over older entries
  • Explicit corrections take precedence over inferred information

Ready to Get Started?

Clone the repository and follow the steps above. For detailed documentation, troubleshooting, and advanced usage, check out the full README on GitHub.