Run the memory in your own account.
Memrain is available as self-hosted, MIT-licensed software. You control the deployment and its stored knowledge, while your chosen AI clients use it through MCP.
Illustrative example
Can we evaluate Memrain with our own project notes?
Deploy a small stack in your AWS account, import a selected set of notes and connect a client with a limited source grant.
Illustrative scenario, not a live product result.A deployment you can inspect.
The documented stack uses an EC2 instance, Postgres with pgvector, EFS and AWS Secrets Manager. Terraform describes the infrastructure, and Docker Compose runs the services. The database holds the canonical page content and search index. Open source gives you access to the implementation and deployment configuration, so your technical team can review what runs and how requests reach it.
Know where information travels.
Stored pages, embeddings and operational logs live in your AWS account. Embeddings and optional model features use Amazon Bedrock. Retrieved passages go to the AI client you connect and its model provider. Default ingress uses Cloudflare Tunnel; the documented Caddy alternative changes that path. Review the selected regions, providers and configuration against your own data requirements.
Own the operating responsibilities.
You run updates, manage credentials, review spending and test restoration. The standard deployment uses one EC2 instance and default single-AZ RDS; it does not provide a high-availability or uptime commitment. Review the setup guide and choose operational responsibilities that fit your deployment.
Keep the context. Move the work forward.
Connect your knowledge to the AI you already use. Find your starting point and make the next conversation count.