Memory for every session.

One store across Claude, ChatGPT, and Codex. The next chat already knows.

Protocol
MCP
Scope
Cross-agent
Data
Yours only
Status
Live

How it works

Connect once. Keep continuity.

Cold starts cost you every day. Shared memory ends that.

1Agent
2MCP
3Memory
4Next
  1. 1. CONNECT

    Add the MCP endpoint. A few guided steps per client.

  2. 2. WORK

    Agents read and write facts, playbooks, and claims.

  3. 3. RETURN

    The next session starts briefed - not blank.

Capabilities

What ships in the store

Click a card to expand

If I had to disconnect it right now, that would be a problem.
Derek Roberts, Founder

Connect

Any MCP client.

Point your agent at one endpoint. Shared memory - not four siloed histories.

Works with

  • Claude
  • Claude Code
  • ChatGPT
  • Codex

Setup is guided at memory.keynodex.com. Browser sign-in. No API key to babysit.

{
  "mcpServers": {
    "keynodex-memory": {
      "url": "https://memory.keynodex.com/mcp"
    }
  }
}

FAQ

Questions

What is an MCP memory server?+

Durable storage over MCP. Your agent reads and writes through standard tools; data outlives the chat.

Which tools work?+

Claude, Claude Code, ChatGPT, and any MCP client. The store is yours - not locked to one model.

What gets stored?+

Facts, playbooks (procedures + history), and claims with verify + freshness. You choose what to save.

Is my data shared?+

No. Isolated per account. We don’t train models on your memories.

Empty on day one?+

It fills as you work. One stack + one recurring fix and the next session is already ahead.

How do I connect?+

Sign in at memory.keynodex.com and add the connector URL. Claude Code: claude mcp add --transport http keynodex-memory https://memory.keynodex.com/mcp

Live now

Stop re-teaching your agents.

Continuity compounds. Cold starts don’t.