Skip to main content
Glama
README.md
# Memwyre MCP Server

[![Glama Score](https://glama.ai/mcp/servers/memwyreai/mcp-memwyre/badges/score.svg)](https://glama.ai/mcp/servers/memwyreai/mcp-memwyre)
[![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)

Universal persistent memory layer and knowledge retrieval engine for AI agents across **Claude Code, Cursor, VS Code, and OpenClaw**.

## Features

- **MCP-Native Persistence**: Capture and recall decisions, architectural notes, and solutions across sessions.
- **Benchmarked 73% LoCoMo Accuracy**: Built on Memwyre's hybrid dense-vector, BM25, and cross-encoder reranking architecture.
- **Cross-Tool Continuity**: Share identical memory across your entire developer toolchain.
- **Fast & Lightweight**: Direct cloud API integration with sub-second retrieval.

## Tools Provided

- `search_memwyre(query: str, limit: int = 5)`: Semantic search over saved memories and notes.
- `save_memory(content: str, tags: list, title: str)`: Ingest facts and solutions into your vault.
- `list_memories(limit: int = 10, skip: int = 0)`: Browse recent entries.
- `get_document(document_id: int)`: Retrieve complete document content.

## Quick Start

### 1. Set your API Key
Generate an API key from your [Memwyre Dashboard](https://app.memwyre.tech/settings).

```bash
export MEMWYRE_API_KEY="bv_sk_your_api_key_here"
```

### 2. Configure Your Client

#### Claude Desktop / Cursor (`mcp.json` / `claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "memwyre": {
      "command": "uvx",
      "args": ["mcp-memwyre"],
      "env": {
        "MEMWYRE_API_KEY": "bv_sk_your_key_here"
      }
    }
  }
}
```

#### Or using Docker:
```json
{
  "mcpServers": {
    "memwyre": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "-e", "MEMWYRE_API_KEY=bv_sk_your_key", "ghcr.io/memwyreai/mcp-memwyre"]
    }
  }
}
```

#### Or connecting directly to our hosted endpoint:
```json
{
  "mcpServers": {
    "memwyre": {
      "url": "https://api.memwyre.tech/mcp",
      "headers": {
        "Authorization": "Bearer bv_sk_your_key_here"
      }
    }
  }
}
```

## License
Apache-2.0

TDQS

A3.7/5.0

Scored across 4 tools

Disambiguation4/5

Each tool has a fairly clear role: search finds relevant memories, list shows recent ones, get retrieves by ID, and save creates new content. Some minor overlap exists between search and list for locating memories, but the descriptions make the distinction workable.

Naming Consistency3/5

Most tools follow a verb_noun pattern (save_memory, list_memories, get_document), but search_memwyre breaks the pattern by using the product name instead of a noun like 'memories'. There is also mild inconsistency between 'memory' and 'document' as the noun for the same underlying resource.

Tool Count5/5

Four tools is a well-scoped size for a personal knowledge vault server. Each tool covers a distinct core operation—save, search, list, and get—without unnecessary bloat.

Completeness3/5

The server covers create and read workflows well, including semantic search, but it lacks update and delete operations for memories or documents. This is a noticeable gap in lifecycle coverage, though simple note-taking use cases can still succeed.

Maintenance

ActivityMaintained
ResponsivenessNo issues