Practical overview
When it fits
Use YantrikDB MCP when an MCP-compatible agent needs persistent memory between sessions, or when memory calls should be routed to a shared HTTP cluster.
Who it is for
- Developers configuring MCP-compatible AI assistants.
- Teams building agents that need memory shared across sessions.
- Builders choosing between local SQLite storage and a remote YantrikDB HTTP cluster.
Documented uses
- Store, recall, correct or forget agent memories.
- Retrieve relevant information even when a question uses different words from the stored memory.
- Link memories in a knowledge graph and identify conflicting stored beliefs.
- Provide an agent with persistent memory through an MCP client.
Requirements & limits
The repository is public and lists the MCP project under the MIT license. It documents installation through pip or uvx and a Claude Code plugin route.
- Local mode stores the database on the machine running the server; the default path is ~/.yantrikdb/memory.db.
- The default local embedder is bundled and 64-dimensional; the README also lists optional 384-dimensional ONNX and 256-dimensional multilingual options.
- Cluster mode requires a YantrikDB HTTP cluster and a database token, and stores memories on the cluster rather than locally.
- The README describes SSE as a legacy mode for a single remote instance and cautions that idle SSE connections can drop.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Configure the local MCP server
Add YantrikDB MCP as a local server using the documented uvx configuration.
Before you start
- An MCP-compatible client
- uv installed and available on PATH
Steps
Add a server entry named yantrikdb to the client's MCP configuration.
Set its command to uvx and its argument to yantrikdb-mcp.
Start or reconnect the client so it can use the configured server.
Expected outcome
The client is configured to run YantrikDB MCP locally; the default local storage path is ~/.yantrikdb/memory.db.
Guide limitations
- The README documents the configuration but does not provide client-specific setup instructions.
- This recipe covers configuration, not a verified client session.
Connect the MCP server to an HTTP cluster
Configure YantrikDB MCP to forward tool calls to an existing YantrikDB HTTP cluster.
Before you start
- A YantrikDB HTTP cluster
- Cluster node URL or URLs
- A database bearer token from the cluster
Steps
Add a yantrikdb server entry to the MCP client's configuration.
Set the command to uvx and the argument to yantrikdb-mcp.
Set YANTRIKDB_SERVER_URL to the cluster node URL or comma-separated node URLs.
Set YANTRIKDB_TOKEN to the database bearer token.
Expected outcome
The MCP server is configured to forward memory tool calls to the cluster, where the README says memories are accessible from any machine.
Guide limitations
- This requires a separately available YantrikDB HTTP cluster and database token.
- The recipe documents configuration only; it does not verify cluster connectivity.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- yantrikos/yantrikdb-mcp · Official homepage title · retained in its source language