Guide · Personal agents

How to share one memory between Claude Code, Codex and Claude Desktop

Connect every agent to the same remote MCP memory server. Built-in memories (CLAUDE.md, Claude Code auto memory, Codex memories) each live inside one tool, and usually on one machine; an MCP server sits outside all of them, so each connected agent reads and writes the same entries.

With Morphika that is one command in Claude Code, two in Codex, and one pasted URL in Claude Desktop. Each connection signs in through your browser, and your personal memory stays private to you.

About Morphika: Morphika is secure, shared memory for personal AI agents like Claude Code, Codex and Claude Desktop. How it is secured.

Why don't my agents already share memory?

Each tool keeps its own. Claude Code reads CLAUDE.md files and writes auto memory under ~/.claude/projects/, which its docs describe as machine-local. Codex reads AGENTS.md, and its optional memories (off by default) live under ~/.codex/memories/. Claude Desktop and claude.ai have their own memory. None of them reads the others' files, so a preference you teach one is unknown to the rest.

How do I connect all three to the same memory?

Claude Code

Terminal

claude mcp add --transport http morphika https://mcp.morphika.ai/mcp --scope user

Then start Claude Code, type /mcp, pick morphika and choose Authenticate. --scope user makes it available in every project on the machine.

Codex

Terminal

codex mcp add morphika --url https://mcp.morphika.ai/mcp
codex mcp login morphika

Claude Desktop or claude.ai

Step 1: Open Settings → Connectors and choose Add custom connector.
Step 2: Name it Morphika and paste https://mcp.morphika.ai/mcp.
Step 3: Click Add, then Connect, and sign in to Morphika in the browser.

Every connection opens the same browser sign-in (OAuth 2.1 with PKCE). There is no API key to copy into a config file.

How do I check it works?

Step 1: In one agent, say: “Save to Morphika: I like short answers with the command first.”
Step 2: Open a new session in a different agent and ask: “Check Morphika: how do I like my answers?”
Step 3: It finds the entry. You can also see it in the dashboard under Memory.

Is my personal memory private?

Yes. Personal memory is visible only to you and to the agents you connect with your account. It saves straight away, with no approval step, because you are its only owner. Every read and write is still recorded in your audit log, so you can see what each agent did.

Morphika also refuses to save anything that looks like a credential, such as an API key or a card number, and masks secrets and personal details in what it returns to an agent.

What if one agent should only read?

Connect it to the read-only address instead: https://mcp.morphika.ai/read/mcp. It gets 7 tools (memory_read, memory_search, memory_list, memory_history, profile_get, document_list, document_read) and cannot write, forget, approve or see the audit log. That is the safe choice for an agent that browses the web or reads untrusted documents.

What about my CLAUDE.md and AGENTS.md files?

Keep them. They are the right place for project rules that belong in the repository and should be reviewed like code. Morphika is for knowledge that crosses tools and projects, such as your preferences, decisions and context you want every agent to have. The two work side by side.

What about Meta Muse, Instinct and other personal agents?

Any agent that can connect to a remote MCP server with OAuth sign-in can use the same memory. Today that means Claude Code, Codex and Claude Desktop. The newer consumer agents are covered in Memory and security for personal AI agents like Muse and Instinct.

Sources

Sources checked September 25, 2026.

Related guides

Published September 25, 2026 · Last updated September 30, 2026