Datavor speaks the Model Context Protocol natively. Any AI tool that supports MCP — Claude, Cursor, Cline, opencode, Gemini, Hermes — can use Datavor's 47 database tools through plain English. Configure once, in 60 seconds.
Anthropic's Claude is Datavor's flagship integration target. MCP was created by Anthropic, and Claude has native support across every surface — desktop app, web app, and the Claude Code CLI. This is the most thoroughly tested combination.
{
"mcpServers": {
"datavor": {
"command": "npx",
"args": ["-y", "datavor"]
}
}
}~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%\Claude\claude_desktop_config.json~/.config/Claude/claude_desktop_config.jsonFor Claude Code (the CLI), add the same server via claude mcp add datavor npx -y datavor.
Cursor speaks MCP natively. Add Datavor once and your editor agent can query schemas, build sync jobs, and run migrations while you code. Particularly powerful for developers who want database operations to live alongside their codebase changes.
{
"mcpServers": {
"datavor": {
"command": "npx",
"args": ["-y", "datavor"]
}
}
}~/.cursor/mcp.json.cursor/mcp.jsonYou can also add it through the UI: Settings → Cursor Settings → MCP → Add new MCP Server.
Cline is a VS Code extension that turns the editor into a fully autonomous coding agent. With Datavor connected through MCP, Cline can execute database tasks as part of larger refactoring jobs — particularly powerful for codebase migrations that involve schema changes.
{
"mcpServers": {
"datavor": {
"command": "npx",
"args": ["-y", "datavor"]
}
}
}Click the Cline icon → ⋯ menu → MCP Servers → Configure MCP Serverscline_mcp_settings.json in your Cline data directoryCline will prompt to approve each Datavor tool call by default — useful safety net for an autonomous agent.
opencode runs entirely in your terminal. It's open source, model-agnostic (works with Claude, GPT, Gemini, local models), and has first-class MCP support. The ideal pairing for developers who live in the terminal and want database work to happen there too — no IDE required.
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"datavor": {
"type": "local",
"command": ["npx", "-y", "datavor"],
"enabled": true
}
}
}~/.config/opencode/opencode.jsonopencode.json in project rootopencode loads project-level config from any opencode.json found while walking up to the nearest Git directory. Note that opencode does not sandbox MCP servers — only enable Datavor in repos you trust.
Google's Gemini family supports MCP through Gemini CLI and Code Assist. Useful if your stack is already in Google Cloud — Datavor's Snowflake and BigQuery connectors pair naturally with Gemini's GCP-aware capabilities.
{
"mcpServers": {
"datavor": {
"command": "npx",
"args": ["-y", "datavor"]
}
}
}~/.gemini/settings.json.gemini/settings.jsonConfirm your Gemini CLI version supports MCP — earlier versions did not. Use gemini --version to check.
Hermes is an autonomous agent framework from Nous Research, built MCP-native from the start. It speaks MCP as both client and server, supports stdio + remote HTTP + OAuth 2.1, and is designed for long-running autonomous workloads — making it a natural pairing for Datavor's pay-per-tool-call agent pricing.
mcp_servers:
datavor:
command: "npx"
args: ["-y", "datavor"]config.yaml or agent-specific .yaml fileHermes also supports remote MCP over HTTP — useful when running Datavor as datavor serve on a separate machine. Combine with Datavor Agent API keys (v3.2) for metered, autonomous data ops.
If it speaks MCP, it works with Datavor. The 6 above are tested and documented. Anything else that implements the MCP client spec should work out of the box — and we'll add it to this list when we've verified.