Engrava MCP server
{
"mcpServers": {
"sovantica-engrava": {
"command": "uvx",
"args": [
"engrava-mcp"
],
"env": {
"ENGRAVA_DB_PATH": "/path/to/directory",
"OPENAI_API_KEY": "sk-..."
}
}
}
}{
"mcpServers": {
"sovantica-engrava": {
"command": "uvx",
"args": [
"engrava-mcp"
],
"env": {
"ENGRAVA_DB_PATH": "/path/to/directory",
"OPENAI_API_KEY": "sk-..."
}
}
}
}{
"mcpServers": {
"sovantica-engrava": {
"command": "uvx",
"args": [
"engrava-mcp"
],
"env": {
"ENGRAVA_DB_PATH": "/path/to/directory",
"OPENAI_API_KEY": "sk-..."
}
}
}
}{
"servers": {
"sovantica-engrava": {
"command": "uvx",
"args": [
"engrava-mcp"
],
"env": {
"ENGRAVA_DB_PATH": "${input:engrava_db_path}",
"OPENAI_API_KEY": "${input:openai_api_key}"
}
}
},
"inputs": [
{
"id": "engrava_db_path",
"type": "promptString",
"description": "Engrava db path",
"password": true
},
{
"id": "openai_api_key",
"type": "promptString",
"description": "Openai api key",
"password": true
}
]
}{
"mcpServers": {
"sovantica-engrava": {
"command": "uvx",
"args": [
"engrava-mcp"
],
"env": {
"ENGRAVA_DB_PATH": "/path/to/directory",
"OPENAI_API_KEY": "sk-..."
}
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall uvx engrava-mcpversion 0.6.0
What it does
Engrava MCP exposes an embedded memory database to AI agents and MCP clients, providing tamper-evident audit, query, and storage capabilities. It is for developers building AI systems that need persistent, searchable memory with full-text search, graph relations, and MindQL query support.
- Store and retrieve agent memories with audit trails to track all changes and maintain integrity.
- Search agent memories using full-text search or semantic search against stored embeddings.
- Query complex memory relationships using MindQL, a query language for the memory graph.
Tools this server exposes
| Tool | Does |
|---|---|
| search_memory | Search the memory database using full-text or semantic search. |
| store_memory | Write new memories to the database. |
| query_memory | Query memories using MindQL to retrieve data from the graph and relations. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.