02 / MODEL CONTEXT PROTOCOL

One memory. Every AI tool.

Connect Memwyre to Cursor, Claude Desktop, VS Code, and Windsurf via the open Model Context Protocol. Centralized memories follow you across every assistant.

Get MCP Keys → Read Configuration

Terminal — npx install-memwyreMCP v1.1

$ npx -y install-memwyre

Memwyre MCP Protocol Handshake:

✔Authenticated: https://server.memwyre.tech/mcp

✔Tools: search_memwyre, save_memory, approve_memory

✔Configured: ~/.cursor/mcp.json & claude_desktop_config

// Ready: cross-tool memory layer active over stdio

>Cursor: "Recalled 4 architectural memories for active session."

The Problem

Every AI conversation starts from zero.

01 . PROBLEM

Context Amnesia

AI clients do not share context. Moving from VS Code to Claude Desktop forces you to re-explain your stack from scratch.

02 . PROBLEM

Repeated Setup Prompts

Developers waste the first 5 prompts of every session repeating setup preferences, workspace rules, and guidelines.

03 . PROBLEM

Siloed Tooling

Valuable insights discovered in Cursor remain trapped in local editor logs and cannot be leveraged elsewhere.

Protocol Specs

The universal protocol for AI context.

Remote HTTP Tunneling (Recommended)

Utilizes mcp-remote to connect your IDE or agent directly to our secure cloud endpoint over JSON-RPC 2.0.

Standardized Tool Suite

Exposes search_memwyre, save_memory, approve_memory, discard_memory, and generate_prompt directly to your LLM.

Dynamic Multi-Session Recall

When your agent queries context, Memwyre performs semantic hybrid search across your vault with under 300ms latency.

~/.cursor/mcp.json

JSON-RPC 2.0

{
  "mcpServers": {
    "memwyre": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://server.memwyre.tech/mcp",
        "--header",
        "Authorization:Bearer bv_sk_your_key_here"
      ]
    }
  }
}

Supported MCP Clients

One command. Infinite context.

Cursor

~/.cursor/mcp.json

Claude Desktop

claude_desktop_config

Windsurf

Cascade MCP provider

VS Code

Extension MCP registry

CLI Installer

npx install-memwyre

Technical Specifications

Technical specifications of our MCP server.

The Model Context Protocol (MCP) server acts as a standard proxy between client IDEs (such as Cursor, VS Code, and Claude Desktop) and your secure Memwyre Cloud vault. Communication is established over standard I/O (stdio) exchanging JSON-RPC 2.0 messages. This ensures that your API tokens remain securely isolated to your local environment.

The server exposes core context management tools to connected LLM agents: search_memwyre for hybrid vector and keyword recall, save_memory to capture insights into the Inbox, approve_memory and discard_memory for curation, and generate_prompt to format retrieved memories into structured prompt templates.

Memwyre provides two connection modes: Remote HTTP Tunneling using the lightweight mcp-remote utility (connecting directly to https://server.memwyre.tech/mcp with TLS 1.3), or full local Python execution running mcp_server.py on Python 3.10+ with local SQLite vector indexing. Both methods ensure seamless cross-client memory portability.

To optimize developer latency, our auto-installer (npx -y install-memwyre) automates OAuth loopback authentication and writes configurations directly to your global client configs. Real-time updates from other devices trigger instant cache invalidations, guaranteeing your agents operate on fresh context.