Connecting 📝MythOS to 📝Cursor gives the AI-native editor direct, identity-aware access to your memos, knowledge graph, and RAG-powered library chat through the 📝Model Context Protocol (MCP).

Cursor is an MCP-compatible client, so any MCP server plugs into it like a first-class tool. MythOS ships its MCP server as the mythos-mcp npm package, distributed over stdio. Once configured, your entire MythOS library becomes a surface Cursor can read from, write to, and reason over — alongside your code, without copy-pasting between tabs. This guide walks through the setup end-to-end: prerequisites, installation, a working configuration, and the workflow patterns that get the most value out of the integration.

Prerequisites

An active MythOS account on the Scholar or Oracle tier (API key generation requires a paid tier)
Cursor installed with MCP support enabled
Node.js 18+ available on your machine (for npx to fetch mythos-mcp)
A MythOS API key generated at mythos.one/settings/api-keys

Installation

Add MythOS to Cursor's MCP configuration. Edit ~/.cursor/mcp.json (global) or .cursor/mcp.json in a specific project:

{
  "mcpServers": {
    "mythos": {
      "command": "npx",
      "args": ["-y", "mythos-mcp"],
      "env": {
        "MYTHOS_API_KEY": "YOUR_API_KEY",
        "MYTHOS_USERNAME": "YOUR_USERNAME"
      }
    }
  }
}

Restart Cursor. The MythOS tools should appear in the MCP panel. The first npx invocation takes a few seconds while the package fetches; subsequent launches are near-instant.

What You Get

Once connected, Cursor has 25 MythOS tools available. The highest-leverage ones in an editor workflow:

get_context — auto-loads your four augmentation memos (Soul, Style, Human, Memory) so Cursor writes in your voice and follows your conventions
search_memos — pull architectural decisions, API keys, past bug notes, or team conventions directly into the editor
read_memo — load full memo content including collaborator notes for rich inline context
chat_with_library — ask 📝RAG questions against your full library; Cursor surfaces the answer inline
create_memo / update_memo — capture a learning or decision mid-session without leaving the editor
list_templates / read_template — use MythOS memo templates (reference, project, person, etc.) from within Cursor

Full tool reference: 📝MythOS MCP — Tools, Resources & Prompts.

Recommended Workflow

The highest-leverage pattern is to let Cursor read from MythOS at session start and write back as you work.

Start of session — ask Cursor "what do you know about this project?" and let it pull the relevant project-context memo. Architectural decisions, deployment rules, and current-state notes load into the conversation before any code is touched
Mid-session — when you hit an issue, "search my library for [pattern]" often surfaces a memo from months ago that saves hours of rediscovery
End of session — "create a memo capturing this decision" ensures the next session — in Cursor, 📝Claude Code, or any other MCP client — inherits today's learning

Troubleshooting

Tools aren't showing up. Cursor caches the MCP tool list at connect time. After editing mcp.json, fully restart Cursor — not just reload the window
Authentication errors. Verify at mythos.one/settings/api-keys that the key is still active and that MYTHOS_USERNAME matches your MythOS account handle exactly
npx failures on first run. If npx -y mythos-mcp fails with a network error, run npm install -g mythos-mcp once outside Cursor, then change the args in mcp.json to ["mythos-mcp"]

FAQ

Does MythOS work in Cursor on Windows and Linux?

Yes. The mythos-mcp npm package runs on macOS, Linux, and Windows wherever Node.js 18+ is installed.

Can I use MythOS in Cursor without an API key?

Not currently. The Cursor integration uses API-key authentication over stdio. OAuth-based connection is available on Claude.ai and Claude Mobile but not yet on Cursor.

Does Cursor see my private memos?

Yes — but only yours. Authentication is scoped per-user via the API key; Cursor cannot access other users' private memos. MythOS's three-tier visibility (public / link / private) is enforced at the server level, not the client.

Does Cursor work alongside Claude Code?

Yes. Both clients connect to the same MythOS backend via MCP. Memos created in one surface are immediately visible in the other. A common pattern is Cursor for in-editor work and Claude Code for multi-file refactors — both see the same library.

Can I use MythOS with Cursor's autonomous agent mode?

Yes. Cursor 3's autonomous agents can call any connected MCP tool, so an agent can search your library, capture learnings back into memos, and operate across your knowledge graph without manual prompting.

📝MythOS MCP — full MCP server reference (tools, resources, prompts, architecture)
📝How to Set Up Claude Code — companion practitioner guide for the terminal-native workflow
📝Cursor — reference memo on the Cursor editor itself
📝Model Context Protocol (MCP) — the standard this integration runs on

I don't use Cursor myself — I work terminal-native in 📝Claude Code with multiple parallel agents, each with full MythOS access. But the pattern scales well for developers who prefer an editor-centric workflow. The point is not which client you pick; it's that your knowledge, conventions, and learnings follow you across every client, every session, every device. MCP is the standard that makes that possible. MythOS is the library that makes it worth turning on.