vplex-memory
v4.3.5
Published
VPLEX Memory MCP Server — stdio proxy with browser-based auth
Maintainers
Readme
vplex-memory
VPLEX Memory MCP Server — persistent cross-session memory for AI coding tools.
Quick Start
Add to your MCP config (.mcp.json, Cursor settings, etc.):
{
"mcpServers": {
"vplex-memory": {
"command": "npx",
"args": ["vplex-memory@latest"]
}
}
}Or via Claude Code CLI:
claude mcp add vplex-memory -- npx vplex-memory@latestSystem Prompt
For best results, add the VPLEX Memory instructions to your project's CLAUDE.md (or equivalent). Copy the contents of SYSTEM_PROMPT.md into your project instructions file.
This tells the AI assistant to automatically use memory tools — searching before decisions, storing after learning, and tracking tasks across sessions.
How It Works
This is a thin stdio proxy that:
- Speaks MCP — JSON-RPC over stdin/stdout
- Authenticates via browser — device auth flow (open URL, approve)
- Caches tokens —
~/.vplex/session.jsonwith 0600 permissions - Proxies to memory-core — all tool calls forwarded to the VPLEX Memory API
Authentication
On first use, the server triggers a device auth flow:
1. Open: https://vplex-dashboard.vercel.app/device?code=ABCD-1234
2. Approve the deviceAfter approval, the token is cached and refreshed automatically.
Environment Variables
| Variable | Default | Description |
|----------|---------|-------------|
| VPLEX_API_URL | https://vplex-memory-v1.vercel.app | Memory Core API URL |
| VPLEX_WEB_URL | https://vplex-dashboard.vercel.app | Web app URL (for auth) |
| VPLEX_PROJECT_PATH | auto-detected | Override project root |
36 MCP Tools
Core Memory: init, store, search, modify, expand, list, export Knowledge Graph: explore, stats, path, invalidate, timeline, at_time, contradictions, consolidate, invalidate_relation Thinking: start_thinking, add_thought Projects: projects, tasks, project_info, project_link Sessions: session_recap, get_rules, list_projects Analysis: insights, upload_document Auth: auth_status, signin, logout Intelligence: postmortem, postmortem_warnings, pattern_feedback, audit_history, code_quality_check Smart Context: smart_context (single-call: auto-detects intent, fetches memories + rules + warnings + graph)
Compatible With
- Claude Code
- Cursor
- Windsurf
- Any MCP-enabled tool
