Engram Alpha
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| ENGRAM_API_KEY | No | Bearer token to protect HTTP / OpenAPI endpoints when exposed. | None / Open Localhost |
| ENGRAM_DB_PATH | No | Absolute filesystem path for the SQLite WAL database. | ~/.engram/engram.sqlite |
| ENGRAM_ALLOWED_ORIGIN | No | Custom CORS allowed origin header for web agents. | * |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| save_memoryA | Save a new memory node with dense vector embedding, category, importance, project namespace, and conflict detection. |
| search_memoryB | Search memory using 4-Way Reciprocal Rank Fusion (RRF): Fuses Dense Vector Cosine Similarity + Trigram FTS5 Lexical + Graph Spreading Activation + ACT-R Decay. |
| extract_and_save_memoryC | Autonomous Memory Extractor Agent: Deconstructs text into atomic facts, extracts entity triples for the knowledge graph using hybrid local LLM sidecar + expanded NLP regex heuristics, and indexes vectors. |
| save_graph_relationC | Save a Subject-Predicate-Object relation with bi-temporal validity and project namespace. |
| query_graphB | Query knowledge graph relations with recursive multi-hop path traversal and bi-temporal filtering. |
| deduplicate_memoriesA | Autonomous Memory Deduplication & Semantic Merging Agent: Finds clusters of duplicate/near-identical memory nodes, merges access counts and edges, and prunes redundant duplicate records in chunks of 1,000 nodes. |
| visualize_graphB | Knowledge Graph Topology Visualizer: Generates Mermaid.js and ASCII relational network diagrams for power users. |
| consolidate_reflectionsC | Autonomous Memory Reflector Agent (Episodic Reflection): Synthesizes low-level episodic nodes into durable high-level insights. |
| ingest_obsidianB | Ingest an entire Obsidian markdown vault into the knowledge graph and vector store. |
| checkpoint_dbB | Execute WAL flush, vacuum, and performance optimization. |
| auto_contextA | Auto-Context Boot Tool for Agents: Recalls top high-importance active memories formatted in XML for session initialization. |
| edit_memoryA | Edit an existing memory's content, importance, or category by its node ID. |
| delete_memoryA | Delete a memory node and cascade-delete all its associated knowledge graph edges by ID. |
| list_memoriesB | List recent memory nodes in the database, ordered by importance and recency. |
| get_statsA | Get system statistics, knowledge graph counts, and active hardware acceleration tier. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 15 tools
Most tools have distinct targets: memory nodes, graph relations, ingestion, maintenance, and stats. A few pairs like save_memory vs extract_and_save_memory or search_memory vs auto_context could overlap, but their descriptions clarify the intended workflow. No tools are truly redundant.
Tool names are consistently lowercase snake_case and mostly follow a verb-first pattern. The main outlier is auto_context, which is a noun phrase, and graph-related names mix save_graph_relation, query_graph, and visualize_graph. Overall the pattern remains predictable and readable.
15 tools is at the upper edge of ideal but well-scoped for a memory system that combines vector storage, knowledge graph operations, ingestion, reflection, and maintenance. Each tool serves a distinct purpose, and none feel purely decorative or redundant.
Memory CRUD is well covered with save, search, edit, delete, and list, and graph relations have save and query support. Missing direct graph relation editing/deletion and a get-memory-by-ID tool are minor gaps, but cascade deletion and search/list provide workarounds.