redash-mcp 0.1.0

Model Context Protocol (MCP) server for Redash - manage queries, dashboards, and visualizations

redash-mcp

Model Context Protocol (MCP) server for Redash - manage queries, dashboards, and visualizations through AI assistants like Claude.

Features

• 5 tools, 22 actions - compressed for minimal context usage
• Full query management (list, search, create, update, archive, delete, run, adhoc)
• Dashboard management (list, get, create, publish, delete)
• Widget management (add, delete)
• Visualization creation (pie, line, bar, counter charts)
• Data source listing

Installation

pip install redash-mcp

Or with uvx:

uvx redash-mcp

Configuration

Environment Variables

Variable	Required	Description
REDASH_URL	Yes	Your Redash instance URL (e.g., https://redash.example.com)
REDASH_API_KEY	Yes	Your Redash API key
REDASH_TIMEOUT	No	Request timeout in seconds (default: 30)

Claude Desktop

Add to ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "redash": {
      "command": "uvx",
      "args": ["redash-mcp"],
      "env": {
        "REDASH_URL": "https://your-redash-instance.com",
        "REDASH_API_KEY": "your-api-key"
      }
    }
  }
}

Or if installed via pip:

{
  "mcpServers": {
    "redash": {
      "command": "redash-mcp",
      "env": {
        "REDASH_URL": "https://your-redash-instance.com",
        "REDASH_API_KEY": "your-api-key"
      }
    }
  }
}

Tools

redash_data_sources

List all available data sources.

redash_query

Manage Redash queries.

Action	Parameters	Description
list	page	List all queries (paginated)
search	q	Search queries by name
get	id	Get query details
create	name, query, data_source_id	Create new query
update	id, query?, name?	Update existing query
archive	id	Archive (soft-delete) query
delete	id	Permanently delete query
run	id, timeout?	Execute query and wait for results
adhoc	query, data_source_id	Execute SQL without saving

redash_dashboard

Manage Redash dashboards.

Action	Parameters	Description
list	page	List all dashboards
get	id	Get dashboard with widgets
create	name	Create new dashboard
publish	id	Publish dashboard (remove draft)
delete	id	Delete dashboard

redash_widget

Manage dashboard widgets.

Action	Parameters	Description
add	dashboard_id, viz_id	Add visualization to dashboard
delete	id	Remove widget from dashboard

redash_viz

Create visualizations.

Type	Parameters	Description
pie	query_id, name, x, y	Pie chart
line	query_id, name, x, y, datetime?	Line chart
bar	query_id, name, x, y, stacked?	Bar chart
counter	query_id, name, x, suffix?	Counter/KPI

Note: For multiple Y columns, pass comma-separated values: y="count,total,avg"

Examples

Create a dashboard with visualizations

1. redash_data_sources() → get data_source_id
2. redash_query(action="create", name="Daily Stats", query="SELECT ...", data_source_id=1)
3. redash_viz(type="line", query_id=123, name="Trend", x="date", y="count")
4. redash_dashboard(action="create", name="My Dashboard")
5. redash_widget(action="add", dashboard_id=456, viz_id=789)
6. redash_dashboard(action="publish", id=456)

Run ad-hoc query

redash_query(action="adhoc", query="SELECT COUNT(*) FROM users", data_source_id=1)

Search and update query

redash_query(action="search", q="daily")
redash_query(action="update", id=123, query="SELECT ... WHERE date > NOW() - INTERVAL '7 days'")

Python Library Usage

You can also use redash-mcp as a Python library:

import os
os.environ["REDASH_URL"] = "https://your-redash.com"
os.environ["REDASH_API_KEY"] = "your-key"

from redash_mcp import (
    list_queries, create_query, run_query,
    create_dashboard, publish_dashboard,
    line, bar, pie, counter,
    add_widget
)

# Create query
q = create_query("My Query", "SELECT * FROM events", data_source_id=1)

# Create visualization
viz = line(q["id"], "Events Trend", x="date", y=["count"])

# Create dashboard and add widget
d = create_dashboard("My Dashboard")
add_widget(d["id"], viz["id"])
publish_dashboard(d["id"])

Why redash-mcp?

• Context efficient - Only 5 tools (~400 tokens) with 22 actions
• Full-featured - Queries, dashboards, widgets, and visualizations
• Production ready - Proper error handling and timeouts
• Dual use - Works as MCP server and Python library

License

MIT