Login

Integration

Long-term memory for ChatGPT

ChatGPT's built-in memory is impressive but opaque. You can't see what it remembers, when, or why. With Engram as a custom MCP app you own the bucket, see every memory in your dashboard, and use the same recall layer across Claude.ai, Cursor, and your own scripts.

Get an API key (free) All integrations →

Install

Three steps: sign up for an Engram API key, paste a BYOK LLM-provider key on /models, then drop the snippet below into ChatGPT (web).

Three steps to memory in your agent

  1. Sign up. Free, no card. You'll land on a Getting Started page that walks the next two steps.
  2. Add your LLM key. Engram is BYOK. Paste an OpenAI / Anthropic / Groq / Together / Fireworks key and we'll route every extraction and query call through your provider. You pay your provider directly. We never see your inference.
  3. Paste the snippet below into your agent and restart it. Use Authorization: Bearer <api-key>with the API key from your portal.

OAuth: no config file, no key to paste

ChatGPT handles auth via OAuth. Sign in to Lumetra in the popup and you're done.

  1. In ChatGPT, open Settings → Apps → Advanced Settings, turn Developer Mode on, then click Create App.
  2. Name it Engram and paste this server URL:
    3. MCP server URL
    https://mcp.lumetra.io/mcp/sse
  3. Confirm. ChatGPT opens a popup to lumetra.io. Sign in (or create a free account) and approve access. You'll land back in ChatGPT with the app live.
  4. Open a chat in a model that supports tools (GPT‑5, GPT‑5 Thinking, etc.) and verify with "List my Engram memory buckets."

Heads up: BYOK still required, plus a ChatGPT caching quirk

Engram is BYOK end-to-end. After you're connected, paste an OpenAI / Anthropic / Groq / Together / Fireworks key on the Lumetra portal so store_memory / query_memory don't return 412.

ChatGPT caches its connector list aggressively. If you remove and re-add Engram, log out and back in (or restart the desktop app) before the new entry picks up cleanly. Full walkthrough: Add Engram memory to ChatGPT.

What you can do once memory's wired in

  • Remember your preferred coding conventions and reuse them across every project
  • Recall your partner's favorite restaurants, allergies, and important dates for gift-related questions
  • Pull up the medical history you shared so ChatGPT can reference it across separate chats
  • Capture book / movie / show ratings once and let ChatGPT use them for recommendations later

FAQ

Which ChatGPT plans support custom MCP apps?

Plans with developer-mode connectors: Plus, Pro, Business, Enterprise, and Edu. The path is **Settings → Apps → Advanced settings → Developer mode** (or, for organizations, Workspace Settings → Permissions & Roles → Developer mode).

Why does ChatGPT cache the connector list so aggressively?

It does, especially the desktop app. After removing and re-adding Engram, log out and back in (or restart the desktop app) to pick up the new entry cleanly.

Will it conflict with ChatGPT's built-in memory?

No, they're independent. The agent will draw from both, and you can keep ChatGPT's memory off if you want Engram to be the only source.

Related integrations

Claude.ai (web) OAuth connector: sign in to Lumetra in a popup, with no config file or key paste. Read more → LibreChat Self-hosted ChatGPT alternative. OAuth per LibreChat user (no shared key in librechat.yaml), with a shared-bearer alternative for single-user deploys. Read more → Cursor Cursor desktop editor and the cursor-agent CLI share one MCP config, global or project-scoped. Read more →

Ship durable memory in ChatGPT (web) today

Free tier: 10K memories and 50K retrievals per month. No credit card. Same Engram backend powers all 41 integrations, so memories you write from one client are immediately queryable from the rest.

Start free See pricing