The Model Context Protocol (MCP) is Anthropic's open standard for letting AI assistants call tools. EQQ's MCP Gateway turns every governed query into a tool Claude, Copilot, or Cursor can invoke - without handing the model raw SQL access.
What makes this hard, and how EQQ handles it

- Auth. An API key maps to a role, which maps to a set of queries. The model can only use what that key permits.
- Schema. EQQ exposes each query as an MCP tool with a JSON schema for inputs - the model knows exactly what parameters to pass.
- Safety. Because the model cannot write SQL, prompt injection cannot escalate access.
- Audit. Every AI-triggered run is logged with a marker showing the calling agent.
The six tools EQQ exposes
list_queries- what queries this key can useget_query_detail- columns, description, metadataget_query_parameters- JSON schema of inputsexecute_query- run and return rowsexport_query- write to xlsx/csv/jsonexecute_chart_query- run with chart classification
Under the hood
The gateway is a small .NET 8 service that speaks MCP over HTTP streaming. It calls EQQ's existing REST API with the supplied API key - no back-channel, no special privileges. If the key cannot run a query through the web UI, it cannot run it through MCP either.
Wire it up
{
"mcpServers": {
"eqq2": {
"url": "http://localhost:5000/mcp",
"headers": { "X-API-Key": "YOUR_KEY" }
}
}
}
Add that to your Claude Desktop config, and Claude can query your databases in one conversational turn. See the MCP Gateway docs for the full setup.
Want to see it live? Book a 30-minute demo - we connect EQQ to your database and build a query with you. Book a Demo →