Guides

Use MemHQ from Claude Desktop

Read and write MemHQ memory from inside Claude Desktop and Claude Code via Remote MCP.

Use MemHQ from Claude Desktop

MemHQ runs as a hosted MCP endpoint at https://memhq.ai/mcp. Point Claude Desktop or Claude Code at the URL with your API key and the memhq_add, memhq_search, and memhq_ask tools appear in the tool picker — no install step.

Connect

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "memhq": {
      "url": "https://memhq.ai/mcp",
      "headers": {
        "Authorization": "Bearer mem_..."
      }
    }
  }
}

Restart Claude Desktop. You should see the MemHQ tools in the picker.

Claude Code

claude mcp add memhq --transport http https://memhq.ai/mcp \
  --header "Authorization: Bearer mem_..."

What you can do from inside Claude

Ask Claude to:

  • "Remember that I prefer postgres over mysql" → Claude calls memhq_add.
  • "What did I tell you about my workflow last week?" → Claude calls memhq_search and synthesizes from the results.
  • "Summarize everything you know about my Acme account" → Claude calls memhq_ask against the group graph.

Claude picks the right tool from natural-language phrasing; you don't need to be explicit about which endpoint to call.

Each connected client gets its own session, scoped by the API key it presents. To target a specific external user id, pass user_id in the tool call — otherwise the project default is used.

Self-hosted alternative

Prefer running the server in-process? Install @memhq/mcp-server locally:

npx -y @memhq/mcp-server

…and wire it into your client with command / args instead of url, setting MEMHQ_API_KEY in the env block. Both transports expose the identical tool surface.

Coming soon — full guide

Multi-project switching, group-graph scoping, and recipes for combining the MemHQ MCP with other MCP servers (filesystem, GitHub) are in progress.