Short answer: Export your memories as JSON (Mem0 from its dashboard or export API, Supermemory from Settings in its Console), upload the file to Fulcra, and have an agent turn it into a clean memory file or records. Then point your agents at Fulcra's MCP server or the ChatGPT plugin. Expect to bring over the memory text, not the embeddings or graph links, which are rebuilt by whatever reads them. Fulcra's Free plan includes every feature and 5 GB of storage.
Is switching right for you?
Mem0 and Supermemory are memory APIs: your app calls them to add and search memories for its users. Fulcra is a user-owned context lake: one person's (or one team's) memory, files and data that any agent they connect can read and write.
Switch to Fulcra if:
- The memory belongs to you and your own agents (Claude, ChatGPT, Codex, OpenClaw, Hermes), not to an app's end users.
- You want the same memory in several agents and tools, not one integration.
- You want to see, edit, version and export exactly what's stored.
- You want memory next to your other context: files, notes, calendar, health and location data.
Stay where you are if:
- You run memory for thousands of end users inside your own product and rely on an add/search memory API. That's what Mem0 and Supermemory are built for.
- You depend on their automatic memory extraction, scoring or graph features and don't want an agent managing memory itself.
What carries over and what doesn't
| Carries over | Notes |
|---|
| Memory text | ✅ | The facts, preferences and decisions themselves |
|---|
| User, agent and session IDs | ✅ | Become folders or fields, for example /memory/<agent>/ |
|---|
| Metadata and timestamps | ✅ | Kept in the exported JSON |
|---|
| Embeddings and vector indexes | ❌ | Not portable between systems; not needed to read the memory back |
|---|
| Graph relations (Mem0 graph memory) | ⚠️ | Export includes the facts; relations need rebuilding if you rely on them |
|---|
| API calls in your code | ❌ | Code that calls Mem0's or Supermemory's API needs changing |
|---|
Step 1: Export your memories
From Mem0 (platform):
- In the Mem0 dashboard, open your project's Memory Exports and create an export, then download the JSON.
- Or use the API: call
create_memory_export(), then get_memory_export() to download it. The export runs in the background, so retry if it isn't ready yet. Your API key needs export permission, and exports expire after 7 days.
- For one user at a time,
get_all() with a user_id filter returns that user's memories.
From Mem0 (open source): call get_all() for each user ID and save the results as JSON.
From Supermemory:
- In the Supermemory Console, go to Settings and request a data export. You get a JSON file with your documents and memories (up to 25,000). Only owners and admins can request it, and the download link expires after six hours, so save it right away.
- Or page through your memories and documents with the list endpoints in the Supermemory API.
Step 2: Move the export into Fulcra
- Connect an agent to Fulcra. Paste
https://docs.fulcradynamics.com/agent-get-started.txt into Claude Code, Codex or another agent with a shell and let it set itself up. In ChatGPT, add the Fulcra plugin. In Claude and other MCP clients, add https://mcp.fulcradynamics.com/mcp.
- Upload the raw export so you always have the original:
uvx fulcra-api file upload mem0-export.json /memory/imports/mem0-export.json
- Turn it into memory your agents can use. Ask the agent:
Read /memory/imports/mem0-export.json from Fulcra. Write a clean /memory/MEMORY.md grouped by topic: keep durable facts, preferences and decisions, merge duplicates, and drop anything stale. Keep the original file unchanged.
If you need memories as queryable records instead of a file, ask the agent to create a custom data type (for example "memory") and record each item into it.
- Install the memory skill so agents keep it up to date:
npx skills add fulcradynamics/agent-skills. The fulcra-memory skill writes notes and daily logs as versioned files you can roll back.
Step 3: Point your agents at Fulcra
Tell each agent where its memory lives now:
At the start of each session, read /memory/MEMORY.md from Fulcra and check what changed since your last session. When you learn something lasting, update that file.
Run both systems side by side for a week. When your agents answer from Fulcra without the old memory API, turn the old integration off.
What it costs
- Fulcra: the Free plan includes every feature and 5 GB of storage. See pricing for larger plans.
- Mem0 and Supermemory: both have free tiers and paid plans priced by usage (requests, tokens or searches). Check Mem0's pricing and Supermemory's pricing for current numbers.
- The migration itself: usually under an hour for one person's memories. The export and upload are free; the agent cleanup step uses your normal model usage.
FAQ
- Can I export my memories from Mem0? Yes. Create an export in the Mem0 dashboard (Memory Exports) or with the export API, then download the JSON. Exports expire after 7 days, so save a copy.
- Can I export my data from Supermemory? Yes. Owners and admins can request a JSON export of documents and memories from Settings in the Supermemory Console. The download link expires after six hours.
- Do my embeddings carry over? No. Embeddings are specific to each system and model. The memory text carries over, and that's what agents read.
- Is Fulcra a drop-in replacement for the Mem0 or Supermemory API? No. Fulcra is a context lake that agents read and write through MCP, a CLI, a ChatGPT plugin and an API, not a memory API with the same calls. If your product calls their API for its end users, plan to change that code, or keep using them for that part.
- Can I keep using Mem0 or Supermemory alongside Fulcra? Yes. Some people keep a memory API inside their app and use Fulcra for their own agents' shared memory and files.
- How much does it cost to move? The export and upload are free, and Fulcra's Free plan includes 5 GB. Most personal migrations take under an hour.
Written by Fulcra. We make one of the tools described on this page.