| name | memory-share |
| description | Share project auto memory over LAN or relay |
| user-invocable | true |
| arguments | [{"name":"passphrase","description":"Shared passphrase for encryption (required)","required":true},{"name":"--raw","description":"Send memory files as-is without distillation","required":false},{"name":"--relay","description":"Use WebSocket relay server for remote sharing","required":false}] |
Memory Share
You are sharing the project's Claude Code auto memory with another machine or person.
There are two modes: distilled (default) and raw (--raw).
Distilled Mode (default)
Step 1: Read memory files
Read all files from your auto memory directory (~/.claude/projects/<project>/memory/):
MEMORY.md (always present)
- Any additional topic files (
debugging.md, patterns.md, etc.)
Step 2: Distill the memory
Analyze all memory files and produce a single, well-organized Markdown document:
# Project Memory Distillation
## Project Overview
(Brief description of the project based on accumulated memory)
## Key Patterns & Conventions
- (coding conventions, architectural patterns, etc.)
## Important File Paths
- (critical files and their roles)
## Debugging Insights
- (known issues, solutions, workarounds)
## Workflow Preferences
- (user preferences for tools, testing, deployment)
## Architecture Notes
- (key architectural decisions and their rationale)
Consolidate, deduplicate, and organize the information clearly. Remove session-specific noise and keep only stable, reusable knowledge.
Write the distilled memory to ${CLAUDE_PLUGIN_ROOT}/.tmp/claude-memory-payload.md.
Step 3: Start the sharing server
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/serve.py "{{passphrase}}" ${CLAUDE_PLUGIN_ROOT}/.tmp/claude-memory-payload.md
Tell the user:
- Instruct the receiver to run:
/claude-spread:memory-receive {{passphrase}}
- The server stays open for multiple receivers until Ctrl+C
Raw Mode (--raw)
Step 1: Bundle memory files
python3 ${CLAUDE_PLUGIN_ROOT}/skills/memory-share/scripts/bundle.py ${CLAUDE_PLUGIN_ROOT}/.tmp/claude-memory-payload.json
This bundles all .md files from the auto memory directory into a JSON file.
Step 2: Start the sharing server
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/serve.py "{{passphrase}}" ${CLAUDE_PLUGIN_ROOT}/.tmp/claude-memory-payload.json
Tell the user:
- Instruct the receiver to run:
/claude-spread:memory-receive {{passphrase}}
- The server stays open for multiple receivers until Ctrl+C
Relay Mode (Remote Sharing)
If the user passes --relay, add the --relay flag to the serve.py command:
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/serve.py --relay "{{passphrase}}" ${CLAUDE_PLUGIN_ROOT}/.tmp/claude-memory-payload.md
Tell the user:
- The room code shown in the output
- Instruct the receiver to run:
/claude-spread:memory-receive --relay --room <room_code> {{passphrase}}
- Requires
pip install websockets if not already installed