DB
databox/mcp
LIVE | mcp.databox.com

> black_box_bi

data_in | Databox | insights_out

Turn any MCP-compatible AI into a complete BI pipeline.
No dashboards. No SQL. No setup.

$ connect --now see pipeline
mcp_endpoint
$ MCP_SERVER="https://mcp.databox.com/mcp"
// the_concept

Databox as a black box

Two tools. Complete BI pipeline. Your AI handles the rest.

INPUT
ingest_data

Push any data into Databox. Metrics, events, custom fields - your AI structures and sends it.

// AI sends data
ingest_data({
dataset: "sales",
data: [{revenue: 42000}]
})
DB
Databox

Stores, aggregates, visualizes.
You don't need to touch it.

← black box →
OUTPUT
query_dataset_with_ai

Ask questions in plain English. AI analyzes data, finds patterns, returns insights.

// AI queries data
query_dataset_with_ai({
question: "What drove
revenue growth?"
})
example: automated reporting
user: "Send our Stripe revenue to Databox"
---
AI: Fetches Stripe data, calls ingest_data
# Result: 847 records ingested
example: instant analysis
user: "Why did churn spike last week?"
---
AI: Calls query_dataset_with_ai , analyzes patterns
# Insight: 73% from pricing tier change
// featured_tools

The two tools that matter

ingest_data
data -> databox

Your AI becomes a data pipeline. Any source, any format - structured and delivered to Databox automatically.

+ Push metrics from any API or system
+ AI handles data transformation
+ Batch or real-time ingestion
+ Schema validation built-in
# Example: AI pipelines Stripe to Databox
user: "Sync yesterday's Stripe transactions"
# AI fetches, transforms, ingests
{"success": true, "rows": 234}
query_dataset_with_ai
databox -> insights

Natural language analytics. A ReAct agent that understands your schema, writes queries, and explains what it finds.

+ Schema-aware reasoning
+ Multi-step analysis chains
+ Trend detection & anomaly finding
+ Human-readable explanations
# Example: Natural language analysis
user: "Compare Q1 vs Q2 performance"
# AI analyzes, compares, explains
{"insight": "Q2 +23%, driven by..."}
// all_tools

Complete toolkit

11 tools for full Databox control

ingest_data

Push data to datasets

query_dataset_with_ai

AI-powered analysis

create_dataset

Define new datasets

create_data_source

Create data sources

list_accounts

List accessible accounts

list_data_sources

Browse data sources

list_data_source_datasets

View source datasets

get_dataset_ingestions

Track ingestion history

delete_dataset

Remove datasets

delete_data_source

Remove data sources

// setup

Connect in 60 seconds

Any MCP-compatible client. One endpoint.

SERVER
https://mcp.databox.com/mcp
TRANSPORT
http // standard HTTP transport
AUTHENTICATION
RECOMMENDED
X-API-Key Header
headers: {"X-API-Key": "pak_xxx"}
ALTERNATIVE
Query Parameter
url: ...mcp?api_key=pak_xxx
QUICK SETUP
claude_desktop_config.json 
{
  "mcpServers": {
    "databox": {
      "type": "http",
      "url": "https://mcp.databox.com/mcp",
      "headers": {
        "X-API-Key": "YOUR_DATABOX_API_KEY"
      }
    }
  }
}
path: ~/Library/Application Support/Claude/claude_desktop_config.json
// missing api key?

Get your Databox API key

required for authentication

$ get_api_key