{
"mcpServers": {
"falkordb-dev": {
"command": "npx",
"args": ["-y", "@falkordb/mcpserver@latest"],
"env": {
"FALKORDB_HOST": "dev.falkordb.local",
"FALKORDB_DEFAULT_READONLY": "false"
}
},
"falkordb-prod-replica": {
"command": "npx",
"args": ["-y", "@falkordb/mcpserver@latest"],
"env": {
"FALKORDB_HOST": "replica.falkordb.com",
"FALKORDB_DEFAULT_READONLY": "true"
}
}
}
}
## Frequently Asked Questions
<AccordionGroup>
<Accordion title="What environment variables are required?">
Only `FALKORDB_HOST` and `FALKORDB_PORT` are required (defaulting to `localhost` and `6379`). Username, password, transport mode, and logging are all optional configuration options.
</Accordion>
<Accordion title="What is the difference between stdio and HTTP transport?">
**stdio** is the default mode for direct integration with AI clients like Claude Desktop, communicating via standard input/output. **HTTP** exposes the server over a network port for remote access, supporting API key authentication via Bearer tokens.
</Accordion>
<Accordion title="How do I secure the HTTP transport mode?">
Set `MCP_API_KEY` to a secret value. When configured, clients must include an `Authorization: Bearer <key>` header in their requests. This is optional but recommended for any network-exposed deployment.
</Accordion>
<Accordion title="Can I connect to multiple FalkorDB instances?">
Yes. Configure multiple MCP server entries in your client config, each with different `FALKORDB_HOST` values and settings. For example, have a `falkordb-dev` instance with writes enabled and a `falkordb-prod-replica` in read-only mode.
</Accordion>
<Accordion title="When should I use read-only mode?">
Enable `FALKORDB_DEFAULT_READONLY=true` for replica instances, production environments where accidental writes must be prevented, reporting/analytics dashboards, and multi-tenant setups where certain users should only have read access.
</Accordion>
</AccordionGroup>