Morphika: frequently asked questions

Morphika is secure, shared memory for personal AI agents like Claude Code, Codex and Claude Desktop. It is a remote MCP server: your agents save and recall notes, decisions and context there, a person approves changes to shared memory, and every call is recorded in an audit log you can verify.

Claude Code, OpenAI Codex, and Claude Desktop or claude.ai through a custom connector. Other agents that connect to remote MCP servers with OAuth sign-in should work too. Meta Muse is being tested, and Instinct isn't supported; see personal AI agents like Muse and Instinct.

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

Then type /mcp in Claude Code, pick morphika and choose Authenticate. Full guide.

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

Full guide.

Settings → Connectors → Add custom connector, name it Morphika, and paste https://mcp.morphika.ai/mcp. Full guide.

Yes. Only you and the agents you connect can read it. Joining a workspace shares nothing from your personal memory.

Because an agent can be manipulated by text it reads (prompt injection), and a bad write to shared memory reaches everyone. So every workspace write waits for another member to approve it; nobody can approve their own, and undecided requests expire after one hour. Why this matters.

No. You can approve from the web dashboard. The iPhone app (iOS 27 or later) sends a notification and needs Face ID for every decision.

A record of every call through Morphika, linked by SHA-256 hashes so any later change breaks the chain. Verify it yourself with Verify now in the dashboard.

No. Agents sign in through your browser with OAuth 2.1 and PKCE. There are no keys to store or leak, and you can revoke any agent from the dashboard.

Yes. Connect it to https://mcp.morphika.ai/read/mcp. It gets 7 tools (read, search and list memory, see an entry's version history, read the profile, and find and read uploaded documents) and can't write, forget, approve or see the audit log.

The save is refused before anything is stored or queued, and the audit log records only the type of finding, never the value.

23 on the full-access address: memory, documents, approvals, audit, workspaces and connections. See them all.

Yes. An agent can give an entry an expiry date when it saves it, up to five years ahead. From that moment the entry is gone from every read, list, search and export. Each version has its own expiry, so saving the entry again without one makes it permanent. In a shared workspace, setting an expiry is a change like any other and waits for approval.

Yes. Every entry is versioned, and memory_history lists its versions: when each was written, by whom, its expiry, what it replaced and where it came from. Earlier text is not kept, because saving an entry again replaces its text; the history is a record of the versions, not a copy of each one. An agent can also save a new entry that replaces an older one of its own, which then drops out of lists and search for anyone who can read the new one.

Ask your agent to forget it (the memory_forget tool, one entry at a time). You can forget an entry you wrote, or any entry in a workspace you own. In your personal memory it happens at once; in a shared workspace it waits for another member's approval, unless the entry is private to you. A forgotten entry is hidden everywhere immediately, and its history and search data are deleted once 7 days have passed and that memory is next used. The stored entry itself is erased when your account is deleted or, for a shared workspace entry, when that workspace is deleted. The details.

One reserved entry, profile, for the standing context every session should start with: who you are, how you like to work. Agents read it with profile_get. It is capped at 16 KB and never expires. In a workspace only an owner can change it, and the change waits for another member's approval.

Yes, from the dashboard: Account, then Export my memory. You get one JSON file with every entry in your personal memory (its text, summary, version, dates and where it came from). Workspace memory and uploaded documents are not included, and secrets stay masked. Export is only in the dashboard: no agent can page through your whole memory.

Delete your account in the iPhone app (Settings → Delete account) or the dashboard (Account → Delete account), which asks you to type DELETE and your current password. Your personal memory is erased at once. What is kept, and why.

No. morphika.ai sets no cookies and loads no analytics or third-party scripts. See the Cookie Policy.

Morphika is in a private pilot, invite only, and no pricing has been published. Request an invite, or write to support@morphika.ai with questions.

Trixode Studios Inc., a company in British Columbia, Canada. About us.