Engram Server
Give coding agents durable memory with fact extraction and hybrid retrieval backed by SQLite or Postgres.
Overview
Engram Server is a MCP server for the Build phase that provides durable agent memory with fact extraction, hybrid retrieval, and a temporal graph on SQLite or Postgres.
What is this MCP server?
- Durable memory with fact extraction and hybrid retrieval across sessions
- Temporal graph modeling for time-aware recall
- SQLite or Postgres storage backends
- OCI package ghcr.io/jamjet-labs/engram-server:0.5.0 with stdio transport
- Documented at java-ai-memory.dev from jamjet runtime
- Version 0.5.0 packaged as ghcr.io/jamjet-labs/engram-server:0.5.0
- 2 storage backends: SQLite and Postgres
- stdio MCP transport
Community signal: 14 GitHub stars.
What problem does it solve?
Agents forget everything when the session ends, so solo builders re-explain context and lose decisions between coding sprints.
Who is it for?
Builders creating long-lived agents, personal copilots, or SaaS features that need searchable memory beyond the context window.
Skip if: Simple one-shot codegen tasks with no need to remember user preferences or project history.
What do I get? / Deliverables
After you run Engram and register the MCP server, your agent can store and retrieve structured facts and time-aware relationships across sessions.
- Running Engram MCP server on version 0.5.0
- Persistent fact store with hybrid search tools exposed to the agent
- Time-aware memory graph for multi-session workflows
Recommended MCP Servers
Journey fit
Persistent agent memory is core agent-tooling you add while building products that must remember users, decisions, and context across sessions. Fact extraction, temporal graph, and hybrid retrieval are infrastructure for long-running agents—not a one-off ship or launch task.
How it compares
Durable memory MCP with graph retrieval, not a generic vector-only RAG snippet pack or a single Cursor skill.
Common Questions / FAQ
Who is Engram Server for?
Solo developers and small teams building MCP-hosted agents that need persistent, queryable memory instead of stuffing everything into the prompt.
When should I use Engram Server?
Use it in Build while you design agent-tooling for products that span multiple sessions, users, or long-running tasks.
How do I add Engram Server to my agent?
Run the ghcr.io/jamjet-labs/engram-server:0.5.0 image (or equivalent from the jamjet repo), configure SQLite or Postgres, attach stdio transport, and add the server entry to your MCP client.