MCP

CLOUD

Overview

The Model Context Protocol (MCP) is an open standard that lets AI assistants and agents (such as GitHub Copilot, Claude, and Cursor) securely connect to external systems through a common interface. Instead of relying on static, pre-trained knowledge, an MCP-compatible client can discover and call a well-defined set of "tools" exposed by an MCP server, letting it take real, up-to-date actions against a live system on your behalf.

Dragonfly Cloud provides a hosted MCP server that exposes your Dragonfly Cloud account and data stores as tools for any MCP-compatible AI assistant or agent. Once configured, an AI agent can use these tools to inspect and manage your Dragonfly Cloud resources on your behalf.

Security and Responsibility

MCP tools can perform destructive, irreversible operations on your Dragonfly Cloud account, including modifying or deleting data stores, networks, and connections, depending on the scope of the API key you provide. An AI agent will act on instructions from its underlying model, which can misinterpret a request or be manipulated by malicious or untrusted content it processes. You are responsible for how you configure and operate an MCP client against your Dragonfly Cloud account. In particular:

Control access. Treat your MCP API key like any other credential. Only share it with MCP clients and agents you trust, and store it securely.
Apply least privilege. Prefer the Read-Only MCP scope unless the agent genuinely needs to make changes. Only use Full-Access MCP when read/write access is required, and rotate or revoke keys that are no longer needed.
Review AI agent actions. Review the tool calls an agent proposes before approving them, especially for any operation that modifies or deletes resources, and monitor your account activity afterward.
Scope the blast radius. Use a dedicated API key per agent or integration rather than reusing one key everywhere, so you can revoke access independently if a specific client misbehaves or is compromised.

Dragonfly Cloud does not expose your data store credentials (such as passwords) through the MCP server by default.

Generating an API Key

Access to the MCP server is authenticated using a Dragonfly Cloud API key with an MCP-specific scope:

Navigate to the Access > API Keys tab in Dragonfly Cloud.
Click the +API Key button to open the Create Key dialog.
In the Create Key dialog, enter a meaningful name for the key (e.g., mcp).
In the Permissions dropdown, select one of the following scopes:
- Full-Access MCP — allows the agent to both read and modify your Dragonfly Cloud resources.
- Read-Only MCP — allows the agent to read your Dragonfly Cloud resources, but not modify them.
Click Create, and the dialog will show the created key.
Click Copy API Key to copy the key to the clipboard.
Store the key somewhere secure for later usage. Dragonfly Cloud doesn't store the key itself.

Configuring an MCP Client

The Dragonfly Cloud MCP server is available at https://api.dragonflydb.cloud/v1/mcp, and is accessed over HTTP using your API key as a bearer token in the Authorization header.

Add the following to your MCP client's configuration, replacing <DFCLOUD_API_KEY> with the API key generated above:

{
  "servers": {
    "dragonflycloud": {
      "url": "https://api.dragonflydb.cloud/v1/mcp",
      "headers": {
        "Authorization": "Bearer <DFCLOUD_API_KEY>"
      }
    }
  }
}

note

The exact location and format of this configuration depends on your MCP client. Consult your client's documentation for how to add a remote, HTTP-based MCP server.

Once configured, restart or reload your MCP client so it picks up the new server. Your AI assistant should now be able to discover and use the tools exposed by the Dragonfly Cloud MCP server, scoped to the permissions of the API key you used.

Detailed Client Setup

Choose your preferred AI client below and follow its setup instructions. In each case, replace the placeholder with the Full-Access or Read-Only MCP API key you generated above.

Visual Studio Code

Add the following to your VS Code settings.json file (or to .vscode/mcp.json for a workspace-specific server):

{
  "servers": {
    "dragonflycloud": {
      "url": "https://api.dragonflydb.cloud/v1/mcp",
      "headers": {
        "Authorization": "Bearer <DFCLOUD_API_KEY>"
      }
    }
  }
}

Cursor

Edit your Cursor mcp.json configuration file (Settings > Cursor Settings > MCP) to add the Dragonfly Cloud MCP server:

{
  "mcpServers": {
    "dragonflycloud": {
      "url": "https://api.dragonflydb.cloud/v1/mcp",
      "headers": {
        "Authorization": "Bearer <DFCLOUD_API_KEY>"
      }
    }
  }
}

Claude Desktop and Claude.ai

Claude Desktop and Claude.ai support remote MCP servers through Settings > Connectors:

Open Settings > Connectors and click Add custom connector.
Enter a name (e.g., Dragonfly Cloud) and the server URL https://api.dragonflydb.cloud/v1/mcp.
When prompted for authentication, choose a header-based option and supply the header Authorization with the value Bearer <DFCLOUD_API_KEY>.
Save, then enable the connector in your conversation's tools list.

Claude Code

Add the Dragonfly Cloud MCP server with a single command:

claude mcp add dragonflycloud \
  -t http https://api.dragonflydb.cloud/v1/mcp \
  -H "Authorization: Bearer <DFCLOUD_API_KEY>"

Codex CLI

Edit your ~/.codex/config.toml and include the following:

[mcp_servers.dragonflycloud]
command = "npx"
args = ["-y", "mcp-remote@latest", "https://api.dragonflydb.cloud/v1/mcp", "--header", "Authorization: Bearer <DFCLOUD_API_KEY>"]

LM Studio

In LM Studio's Program > Install > Edit mcp.json panel, add:

{
  "mcpServers": {
    "dragonflycloud": {
      "url": "https://api.dragonflydb.cloud/v1/mcp",
      "headers": {
        "Authorization": "Bearer <DFCLOUD_API_KEY>"
      }
    }
  }
}

Other MCP Clients

Any MCP client that supports remote, HTTP-based servers with custom headers can connect to Dragonfly Cloud using the same two pieces of information:

Server URL: https://api.dragonflydb.cloud/v1/mcp
Header: Authorization: Bearer <DFCLOUD_API_KEY>

Consult your client's documentation for where to enter these values.

Available Tools

The Dragonfly Cloud MCP server exposes the following tools. Tools that create, modify, or delete resources require a Full-Access MCP key; the Read-Only MCP scope is intended for the inspection tools.

Datastores

Tool	Description
`dfcloud_list_datastores`	List all datastores.
`dfcloud_show_datastore`	Show the status and configuration of a datastore.
`dfcloud_create_datastore`	Create a datastore. Provisioning is asynchronous.
`dfcloud_update_datastore`	Update a datastore's name, memory, performance tier, engine settings (cache mode, BullMQ, Sidekiq, Memcached, SSD tiering), network access, backup policy, or maintenance window.
`dfcloud_start_datastore`	Start a stopped datastore.
`dfcloud_stop_datastore`	Stop an active datastore.
`dfcloud_delete_datastore`	Permanently delete a datastore. Irreversible.
`dfcloud_list_datastore_nodes`	List the nodes of a datastore, optionally for a single shard.
`dfcloud_list_datastore_shards`	List the shards of a clustered datastore, with state, slot ranges, and memory sizes.
`dfcloud_list_datastore_migrations`	List slot migrations between shards of a cluster datastore (`active`, `finished`, or `all`).

Monitoring and Diagnostics

Tool	Description
`dfcloud_datastore_info`	Run `INFO` on a datastore (or a specific node) and return JSON.
`dfcloud_datastore_metrics`	Fetch instant Prometheus metrics (CPU, memory, ops/sec, hit rate, connected clients, etc.).
`dfcloud_datastore_resource_usage`	Query CPU and/or memory usage time series over a time range (defaults to the last hour).
`dfcloud_datastore_slowlog`	Run `SLOWLOG GET` and return the slow query log entries.
`dfcloud_datastore_objhist`	Run `DEBUG OBJHIST` and return the object histogram.
`dfcloud_query_audit_log`	Query the audit log, filtered by resource, resource type, event type, operator type, and time range.

Running Commands

Tool	Description
`dfcloud_run_readonly_datastore_command`	Run an allow-listed read-only command (for example `GET`, `SCAN`, `HGETALL`, `INFO`, `MEMORY USAGE`). Commands that can write or modify data, configuration, or server state are rejected.
`dfcloud_run_datastore_command`	Run any command on a datastore, including commands that modify or delete data.

Both tools accept an optional db_index (clustered datastores only support database 0) and an optional node_id to target a specific node.

caution

Command output is data stored by the datastore's own clients. Treat it as untrusted and never let an agent follow instructions found in it.

Backups

Tool	Description
`dfcloud_list_backups`	List backups, optionally filtered by datastore.
`dfcloud_show_backup`	Show the status and details of a backup.
`dfcloud_create_datastore_backup`	Trigger an immediate manual backup, with a retention of up to 30 days (7 by default).
`dfcloud_import_backup`	Import a backup from HTTP(S) or cloud storage URLs so it can later be restored.
`dfcloud_backup_restore`	Restore a backup into an existing datastore. Replaces the datastore's current data.
`dfcloud_delete_backup`	Delete a backup. Irreversible.

Networks and Connections

Tool	Description
`dfcloud_list_networks`	List private networks.
`dfcloud_show_network`	Show the status and configuration of a network.
`dfcloud_create_network`	Create a private network in a provider and region (CIDR `/16` to `/24`).
`dfcloud_delete_network`	Delete a network. Irreversible; remove its connections first.
`dfcloud_list_connections`	List VPC/VNet peering connections.
`dfcloud_show_connection`	Show the status of a connection.
`dfcloud_create_connection`	Create a VPC peering connection between a Dragonfly Cloud network and your own VPC (AWS, GCP) or VNet (Azure).
`dfcloud_delete_connection`	Delete a connection. Irreversible.

Regions, BYOC, and Billing

Tool	Description
`dfcloud_list_regions`	List cloud providers and regions, with availability zones and performance tiers.
`dfcloud_list_instance_families`	List instance families available for the `byoc` performance tier.
`dfcloud_list_byoc_accounts`	List Bring Your Own Cloud (BYOC) accounts.
`dfcloud_show_byoc_account`	Show the status and configuration of a BYOC account.
`dfcloud_list_invoices`	List invoices, filtered by status and date range.

note

Many mutating tools are asynchronous: they return once the operation starts, and the resource reports a transitional status (such as pending, updating, or restoring) until it finishes. Datastore credentials are never returned by these tools; retrieve them from the Dragonfly Cloud dashboard.

Overview​

Security and Responsibility​

Generating an API Key​

Configuring an MCP Client​

Detailed Client Setup​

Visual Studio Code​

Cursor​

Claude Desktop and Claude.ai​

Claude Code​

Codex CLI​

LM Studio​

Other MCP Clients​

Available Tools​

Datastores​

Monitoring and Diagnostics​

Running Commands​

Backups​

Networks and Connections​

Regions, BYOC, and Billing​