Bring every memory with you.
Already on Mem0, Zep, Pinecone or a pgvector table? Import what your agents already know instead of starting from zero, then give it branches, a graph and an MCP server.
Nine importers, one flow.
- Migration guideMem0
Mem0 platform memories
- Migration guideZep
Zep Cloud session memory
- Migration guidePinecone
A Pinecone index
- Migration guideQdrant
A Qdrant collection
- Migration guideChroma
A self-hosted Chroma collection
- Migration guideWeaviate
A Weaviate class
- Migration guideMongoDB
A MongoDB or Atlas collection
- Migration guidePostgres
A Postgres table, with or without pgvector
- Migration guideJSON or CSV
A JSON, JSONL or CSV export from any memory tool
Preview first. Nothing lands until you say so.
Connect
Preview
Map
Run
Import questions.
Will my existing embeddings be reused, or does everything get re-embedded?
Reused when they’re compatible with the destination. Importing straight into a Studio database always works regardless of embeddings, because Axonio’s engine embeds content itself, the same as anything created directly in Studio. Importing into a raw vector collection instead requires the source’s embeddings to already match that collection’s dimensions.
Are my credentials safe? Do I have to share an API key?
Not necessarily. Where a provider has its own export (mem0’s dashboard, for one), the wizard points you to it directly so you can download a file and skip sharing a key entirely. When a live connection is the only option, or simply the one you prefer, a Postgres connection string or API key is encrypted at rest before it’s stored, and a connection is tested before it’s ever saved. You can revoke or delete a saved connection at any time.
What if my provider isn’t mem0, Zep, or Postgres-based?
Export it to JSON, NDJSON, or CSV and use the file import path. It covers any provider that can produce a file, with the same field-mapping and preview steps as every other source.
What happens if a large import fails partway through?
Progress is checkpointed after every batch, so retrying an import resumes from where it stopped instead of starting over and re-writing records that already landed.