/ PRODUCTS / MCP

[1/3]

One memory.
Every AI tool.

Connect Memwyre to Claude, Cursor, VS Code, Windsurf, and any MCP-compatible client. Your context follows you everywhere.

Get MCP Keys → Read Configuration

Terminal — memwyre-mcp

$ npx -y mcp-remote https://server.memwyre.tech/mcp --header "Authorization:Bearer YOUR_KEY"

Initializing Memwyre Model Context Protocol Server...

✔ Fetched auth credentials from environment

✔ Configured schemas for Cursor & Claude Desktop

✔ Listening on stdio protocol stream

// Ready! Ask Cursor: "What did I do yesterday?"

> AI: According to your memory log, you updated the auth flow config...

The Disconnect

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 explain your stack all over again.

02 . PROBLEM

Repeated Explanations

Spend the first 5 prompts of every session typing out setup preferences, workspace configurations, and style guides.

03 . PROBLEM

Scattered Knowledge

Conversations remain locked within client histories. Your team cannot reuse the insights gained in Cursor.

Protocol Specs

The universal protocol for AI context.

Expose context via Stdio

The Memwyre MCP server launches as a subprocess communicating via JSON-RPC 2.0 over standard I/O.

Declare Tools & Prompts

Exposes tools like save_memory, search_memwyre, and generate_prompt directly to your LLM sessions.

Dynamically Retrieve

When your AI agent needs context, it queries your Vault via vector and keyword search in under 300ms.

cursor.config.jsonJSON

{
  "mcpServers": {
    "memwyre-vault": {
      "command": "npx",
      "args": [
        -y,
        mcp-remote,
        https://server.memwyre.tech/mcp,
        --header,
        Authorization:Bearer bv_sk_839f201...
      ]
    }
  }
}

Ecosystem

One command. Infinite context.

Cursor

Auto-inject context

Claude Desktop

Direct agent tool access

Windsurf

Integrated cascade memory

VS Code

Extension terminal sync

Command Line

Local npx command pipeline

Protocol Specs

Technical specifications of our MCP server.

The Model Context Protocol (MCP) server acts as a local proxy between client interfaces (such as Cursor, VS Code, and Claude Desktop) and your secure Memwyre Cloud vault. The communication is established as a stdio subprocess, exchanging JSON-RPC 2.0 messages over standard input and standard output streams. This guarantees that your connection credentials remain private and isolated to your local environment.

The server exposes two primary tools to the LLM agent: save_memory and search_memwyre. When an agent runs a query, it dynamically assesses whether to call these tools. For example, if you ask Cursor "What did we configure yesterday?", the LLM generates a tool call to search_memwyre with semantic query parameters. The local server processes the request, performs a vector search against the Memwyre vault, and returns matching context blocks in less than 300 milliseconds.

Similarly, when you compile a package or explain a new requirement, the agent calls the save_memory tool to persist this context. The local server routes this data back to your cloud vault, ensuring that the next time you open VS Code, Claude Desktop, or Windsurf, the agent starts with the latest context. This removes the stateless amnesia problem and guarantees consistent developer guidelines across all AI client sessions.

To further optimize client query performance, the local MCP proxy maintains a lightweight cache of recently accessed semantic rule blocks. This ensures that repeated lookups of static project parameters do not require a round-trip to the cloud server, reducing latency to single-digit milliseconds. When you make changes to your cloud vault via the browser or the extension, a WebSocket invalidate signal is broadcast to clear this local cache, maintaining a fully synchronized state across your tools. This guarantees real-time updates and minimal latency.

Context infrastructure for AI agents.

Connect your team's stack to a shared memory. Stop repeating settings and setup parameters.

Connect Your AI Tools

One memory.Every AI tool.