
yantrikdb-server
by yantrikos·★ 143·Score 50
Cognitive memory database for AI agents that consolidates duplicates, detects contradictions, and fades stale memories via temporal decay.
Overview
YantrikDB is a unique cognitive memory engine that manages memories rather than just storing them. Unlike traditional vector databases, it implements three key features: forgetting (via temporal decay with configurable half-life), consolidation (collapsing similar memories into canonical forms), and conflict detection (identifying contradictions between memories). It offers significant token savings compared to file-based memory systems - up to 99.9% at 5,000 memories while improving recall precision. The system can be deployed as a library, HTTP server, or MCP server, with proactive triggers that surface what needs attention rather than requiring agents to constantly query.
Try asking AI
After installing, here are 7 things you can ask your AI assistant:
When to choose this
Choose YantrikDB when building AI agents that need sophisticated memory management beyond simple retrieval, especially when dealing with long-term memory, conflict detection, or temporal relevance.
When NOT to choose this
Not ideal if you need simple key-value storage without cognitive features; AGPL license may be problematic for commercial use without compliance.
Tools this server exposes
12 tools extracted from the READMErememberStore new memories with text, importance, and temporal decay
recallRetrieve relevant memories based on query and context
forgetRemove specific memories from the database
correctUpdate or correct existing memories
thinkPerform cognitive operations: consolidate memories, detect contradictions
memoryAccess or manipulate memory structures
triggerGet proactive insights from memory system
graphAccess the knowledge graph of entities and relationships
categoryOrganize or query memories by categories
conflictList or resolve contradictions in stored memories
sessionManage session-specific memories and context
temporalQuery memories based on time: stale, upcoming, or time-bound
Comparable tools
Installation
Quick Start with MCP
pip install yantrikdb-mcpAdd to your MCP client config (typically ~/.claude.json or .mcp.json):
{
"mcpServers": {
"yantrikdb": {
"command": "yantrikdb-mcp"
}
}
}For shared memory across machines:
{
"mcpServers": {
"yantrikdb": {
"command": "yantrikdb-mcp",
"env": {
"YANTRIKDB_SERVER_URL": "http://node1:7438,http://node2:7438",
"YANTRIKDB_TOKEN": "ydb_your_database_token"
}
}
}
}As a Network Server
docker run -p 7438:7438 ghcr.io/yantrikos/yantrikdb:latestFAQ
- How does YantrikDB differ from vector databases like Pinecone or Weaviate?
- Vector databases focus on nearest-neighbor lookup but lack memory management features. YantrikDB implements temporal decay, consolidation of similar memories, and conflict detection - making it a cognitive memory system rather than just a storage layer.
- Can I use YantrikDB without an internet connection?
- Yes, YantrikDB can run as an embedded library that works offline with a local SQLite file. It only requires network connectivity if you use the clustered deployment option.
- How much memory does YantrikDB use at scale?
- Despite storing thousands of memories, YantrikDB maintains a constant ~70 tokens per query, regardless of memory count. This contrasts with file-based systems that grow linearly with memory count.
On Hacker News
Recent discussion from the developer community.
- Story by pranabsarkar · 2026-04-14
Compare yantrikdb-server with
Last updated · Auto-generated from public README + GitHub signals.