Engram
Give long-running agents hierarchical memory—buffer, working, and core layers with decay and promotion—via MCP against an Engram server.
Overview
Engram MCP is a Build-phase MCP server that exposes hierarchical agent memory with buffer, working, and core layers plus decay and promotion.
What is this MCP server?
- Three-layer memory model: buffer, working, and core
- Decay and promotion rules move durable facts into long-term core
- MCP bridge (engram-rs-mcp) talks to a remote Engram service
- Authenticated via ENGRAM_URL and ENGRAM_API_KEY
- npm engram-rs-mcp v0.10.0 over stdio
- Three memory layers: buffer, working, and core
- MCP package version 0.10.0 identifier engram-rs-mcp on npm
- Repository github.com/kael-bit/engram-rs
Community signal: 25 GitHub stars.
What problem does it solve?
Stateless agents forget project context between sessions unless you bolt on ad-hoc notes or bloated prompts.
Who is it for?
Solo builders shipping personal or customer-facing agents that need structured long-term memory beyond a single thread.
Skip if: Simple one-shot codegen tasks, teams without an Engram server to target, or products satisfied with document RAG only.
What do I get? / Deliverables
Your agent can store and retrieve layered memories through Engram so durable facts promote to core while ephemeral detail decays.
- MCP tools backed by three-layer Engram memory
- Decay and promotion behavior for durable agent recall
- Versioned npm bridge engram-rs-mcp to remote Engram API
Recommended MCP Servers
Journey fit
Persistent agent memory is core Build agent-tooling infrastructure for products that must remember context across sessions. Agent-tooling is the canonical shelf because Engram is memory substrate for agents, not a one-off API or launch marketing task.
How it compares
Hierarchical memory service MCP, not a generic vector-database browser or brainstorming methodology skill.
Common Questions / FAQ
Who is Engram MCP for?
Agent builders who run an Engram memory server and want Claude Code, Cursor, or similar clients to read and write layered agent memory over MCP.
When should I use Engram MCP?
Use it during Build when your agent product needs session-spanning recall with buffer, working, and core tiers instead of flat chat history.
How do I add Engram MCP to my agent?
Deploy or use an Engram server, install engram-rs-mcp, configure stdio MCP with ENGRAM_URL and ENGRAM_API_KEY, then expose memory tools to your agent.