@iflow-mcp/kael-bit-engram-rs
v0.14.0
Published
MCP server for engram — hierarchical memory engine for AI agents
Downloads
19
Readme
engram-rs-mcp
MCP server for engram — persistent, brain-like memory for AI agents.
Works with Claude Code, Cursor, Windsurf, OpenClaw, and any MCP-compatible editor.
Quick Setup
{
"mcpServers": {
"engram": {
"command": "npx",
"args": ["-y", "engram-rs-mcp"],
"env": {
"ENGRAM_URL": "http://localhost:3917",
"ENGRAM_API_KEY": "",
"ENGRAM_NAMESPACE": ""
}
}
}
}Claude Code
claude mcp add -s user engram -- npx -y engram-rs-mcpCursor / Windsurf
Add the config above to .cursor/mcp.json or equivalent.
Environment Variables
| Variable | Default | Description |
|----------|---------|-------------|
| ENGRAM_URL | http://localhost:3917 | engram server URL |
| ENGRAM_API_KEY | — | Bearer token for auth |
| ENGRAM_NAMESPACE | — | Namespace for project isolation |
Tools
| Tool | Description |
|------|-------------|
| engram_store | Store a memory with optional importance, tags, kind |
| engram_recall | Hybrid search with budget, time filters, reranking |
| engram_resume | Session recovery — core + recent + topic index + triggers |
| engram_topic | Drill into topic clusters by ID from the knowledge index |
| engram_recent | List recent memories by time window |
| engram_search | Quick keyword search |
| engram_extract | LLM-powered text → structured memories |
| engram_consolidate | Run promotion/decay/merge cycle |
| engram_triggers | Pre-action safety recall |
| engram_stats | Layer counts and status |
| engram_health | Health check |
| engram_delete | Delete a memory by ID |
| engram_update | Update importance, tags, or layer |
| engram_trash | List soft-deleted memories |
| engram_restore | Restore from trash |
| engram_facts | Insert fact triples |
| engram_repair | Fix FTS index + backfill embeddings |
What is engram?
Persistent memory organized along two dimensions — time (three-layer lifecycle: Buffer → Working → Core) and space (self-organizing topic tree). Automatic decay, promotion, dedup, LLM-powered quality gates, and topic clustering. Single binary, ~9 MB.
See the main repo for full documentation.
License
MIT
